Introduction

The development of specifically targeted drugs for hepatitis C virus (HCV), termed direct-acting antivirals (DAAs), has opened a new era in the treatment of chronic hepatitis C [1]. While a number of DAAs have been approved and are now in development, their therapeutic targets are limited to three non-structural proteins of HCV, including non-structural protein 3/4A (NS3/4A) protease, non-structural protein 5A (NS5A) replication complex and non-structural protein 5B (NS5B) polymerase. A combination of the NS3/4A inhibitor asunaprevir and NS5A inhibitor daclatasvir received the first global approval as an all oral interferon-free ribavirin-free regimen for genotype 1 HCV infection in 2014 in Japan [2]. This combination is known to have limited efficacy in genotype 1a patients but is known to be effective in genotype 1b patients [3]. A Japanese phase III trial achieved 84.7 % sustained virologic response (SVR) in interferon-ineligible/-intolerant naïve patients and in previous interferon non-responders with HCV genotype 1b infection [4]. An international phase III trial has also achieved a similar SVR rate for genotype 1b patients [5]. An important baseline factor associated with the therapeutic outcomes of this therapy is the pre-existing NS5A variant at the L31 and/or Y93 positions. In the Japanese trial, pretreatment L31M, Y93H or linked L31V plus Y93H variants were detected in 7, 29 and 1 of the 214 patients with available baseline NS5A population sequences, respectively, and an SVR rate of 40.5 % was achieved among those patients, while without variants, the SVR rates were over 90 % [4]. Importantly, most patients who failed to achieve SVR developed resistance-associated variants (RAVs) at the NS5A Y93 plus L31 positions combined with the NS3/4A D168 position, although such multiple variants rarely exist before the therapy. The NS5A L31 and Y93 variants were reported to persist for a long time, while the NS3/4A D168 variant tends to disappear with time after therapy [6]. Whether these treatment-emergent multiple variants respond to next generation DAA therapies is not clear. One of the most promising regimens for genotype 1 infection is a combination of the NS5A inhibitor ledipasvir and the NS5B polymerase inhibitor sofosbuvir [7–9], which has been approved by the US Food and Drug Administration and will soon be approved in Japan. This regimen achieved over 95 % SVR, and, importantly, the preexisting NS5A variant did not affect therapeutic outcomes in phase III studies. These observations clearly indicate that the ledipasvir/sofosbuvir regimen is quite effective for naturally occurring NS5A variants, but whether RAVs emerging after asunaprevir/daclatasvir therapy respond to the ledipasvir/sofosbuvir regimen is not fully understood.

In the present study, we use human hepatocyte chimeric mice [10] to examine the efficacy of a combination of ledipasvir and a sofosbuvir-like NS5B polymerase inhibitor (GS-558093) for RAVs emerging after asunaprevir/daclatasvir therapy and demonstrate that the NS3/4A and NS5A triple mutant virus is relatively resistant to ledipasvir/GS-558093. The triple mutant virus was clearly suppressed by GS-558093 combined with NS3/4A protease inhibitor, which does not share cross-resistance with asunaprevir. The results suggest that scrupulous attention is needed when applying the same class of DAA on re-treatment for patients with previous DAA failure.

Materials and methods

Human serum samples

Human serum containing genotype 1b HCV was collected from two patients with chronic hepatitis C after obtaining written informed consent under the approval of Osaka University Hospital Ethics Committee. Serum samples were divided into aliquots and frozen at –80 °C until use.

HCV infection to chimeric mice

Humanized liver chimeric mice (Suppl. Materials and Methods) [10], whose chimeric rate of the liver was estimated as over 40 %, were injected intravenously with 100 µl of HCV-positive human serum samples. After inoculation, their blood was collected from an external jugular vein every 1–4 weeks. The HCV RNA levels were measured by the COBAS TaqMan HCV test (Roche Diagnostics, Basel, Switzerland) in 100-fold diluted serum with a lower measurement range of 3.2 log IU/ml serum. All mice studies were conducted in strict accordance with the Guide for the Care and Use of Laboratory Animals from Osaka University Medical School and Central Institute for Experimental Animals (CIEA). All experimental protocols were approved by the Animal Care and Use Committee of Osaka University Medical School and the Animal Care Committee of CIEA.

Treatment of HCV-infected mice with DAAs

After serum levels of HCV RNA reached plateau levels, mice were administered orally once a day for 4 weeks with one of the following: 40 mg/kg of asunaprevir plus 30 mg/kg of daclatasvir, 15 mg/kg of ledipasvir plus 50 mg/kg of GS-558093 (formerly known as PSI-353661, a nucleotide analog inhibitor of NS5B polymerase) [11] and 50 mg/kg of GS-558093 plus 400 mg/kg of telaprevir (Mitsubishi Tanabe Pharma Corp., Osaka Japan). Asunaprevir and daclatasvir were provided by Bristol-Myers Squibb (New York, NY). Ledipasvir and GS-558093 were provided by Gilead Sciences Inc. (Foster City, CA).

Detection of resistance-associated variants of HCV

HCV RNA was extracted from 10 µl of serum samples using QIAamp® Viral RNA Mini (QIAGEN, Hilden, Germany) and reverse transcribed with HCV specific whole reverse primer (Sup. Table 1) using SuperScriptTMIII Reverse Transcriptase (Thermo Fisher Scientific Inc., MA). NS3/4A, NS5A and NS5B regions were amplified using KOD-Plus-Neo (Toyobo, Osaka Japan) with outer primer pairs (Suppl. Tables 1, 2). For population sequencing, the first amplicons were further amplified using KOD-Plus-Neo with inner primer pairs (Suppl. Tables 1, 2). After purification with agarose gel electrophoresis, nucleotide sequences of second amplicons were determined using the BigDye terminator v3.1 cycle sequencing kit (Thermo Fisher Scientific Inc.) by standard Sanger sequencing. For deep sequencing, the first amplicons were further amplified using Platinum PCR SuperMix High Fidelity (Thermo Fisher Scientific Inc.) with inner primer pairs, which attached the adaptor and barcodes (Suppl. Tables 1, 2). After purification with AMPure (Beckman Coulter, CA), the second amplicons, whose nucleotide sizes were shorter than 400 base pairs, were sequenced by the Ion Personal Genome Machine (Ion PGM) Sequencer (Thermo Fisher Scientific Inc.).

Results

We used sera from two patients with chronic HCV genotype 1b infection (Table 1). One was a 63-year-old female who had received asunaprevir/daclatasvir therapy and experienced on-treatment viral breakthrough, while the other was a 64-year-old male who had not received any DAA therapy. Population sequence analysis revealed that the asunaprevir/daclatasvir-experienced patient carried RAVs, NS3/4A D168V and NS5A L31V and Y93H, in association with viral breakthrough, which were not detected before the therapy, while the DAA-naïve patient did not show any substitutions at these three positions. In the former, RAVs at these three positions were detected 2 years after the therapy. Given the deep sequencing data revealing an extremely high frequency of each substitution (D168V 95.0 %; L31V 99.3 %; Y93H 99.3 %), it is reasonable to think that these data did not represent a mixture of viruses with a single substitution at each position, but rather resulted predominantly from viruses carrying simultaneous substitutions at these positions.

Table 1 HCV NS3/4A and NS5A RAVs in the patient sera
Full size table

TK-NOG mice transplanted with human hepatocytes [10] were inoculated with either serum. Mice that developed persistent infection exhibited serum HCV RNA levels at 6–7 log IU/ml in each group (Fig. 1). Although all serum samples (8 from wild-type HCV-infected mice and 7 from mutant HCV-infected mice) were applied to determine the HCV sequence, two of the data of both groups were missing because PCR amplification was unsuccessful (Table 2). Population sequencing revealed that mice inoculated with DAA-naïve sera carried the wild-type virus, while mice inoculated with sera from the asunaprevir/daclatasvir-experienced patient carried the NS3/4A D168V and NS5A L31V and Y93H variants. Deep sequence analysis confirmed high frequencies of substitution at these three positions, although the frequency somewhat differed among individual mice especially at the NS3/4A D168 position.

Fig. 1
figure 1

Serum levels of HCV RNA in human hepatocyte chimeric mice after infection with HCV. Closed circle with solid line demonstrates HCV RNA levels in mice inoculated with wild-type HCV from patient 2. Open circle with dashed line demonstrates HCV RNA levels in mice inoculated with mutant HCV from patient 1. Mean ± standard division were presented. N = 7 or 8 per each

Full size image
Table 2 HCV NS3/4A and NS5A RAVs in mice prior to DAA treatment
Full size table

Mice in each group that developed persistent HCV infection were divided into two treatment groups. One group received 4 weeks of asunaprevir/daclatasvir treatment and the other group received 4 weeks of ledipasvir/GS-558093 treatment. Asunaprevir/daclatasvir therapy and ledipasvir/GS-558093 therapy rapidly deceased serum HCV RNA levels to below the sensitivity, and they were not detected after completion of the therapy except for two mice in the ledipasvir/GS-558093 group (Fig. 2a). In contrast, for mice infected with mutant virus, asunaprevir/daclatasvir therapy reduced serum HCV RNA levels only by 1–2 log IU/ml from baseline, while ledipasvir/GS-558093 therapy reduced serum HCV RNA levels to a slightly larger extent but failed to achieve end-of-treatment response (Fig. 2a). These results suggest that ledipasvir/GS-558093 therapy is relatively ineffective against triple mutant virus compared with wild-type virus. Although mice infected with triple mutant virus failed to eradicate HCV by asunaprevir/daclatasvir therapy or ledipasvir/GS-558093 therapy, no additional substitutions at the NS3/4A, NS5A and NS5B regions were detected after the therapies (Table 3).

Fig. 2
figure 2

Serum levels of HCV RNA in HCV-infected chimeric mice after indicated DAA treatment. a HCV-infected chimeric mice were treated with asunaprevir (ASV)/daclatasvir (DCV) or ledipasvir (LDV)/GS-558093 (NS5B NI) for 4 weeks. Closed circles with solid lines indicate HCV RNA levels in mice infected with wild-type HCV from patient 2. Open circles with dashed lines indicate HCV RNA levels in mice infected with mutant HCV from patient 1. b Mutant HCV-infected chimeric mice that had received 4 weeks of ASV/DCV therapy (closed triangle) or LDV/NS5B NI therapy (open triangle) as in A were treated with telaprevir (TVR)/NS5B NI for 4 weeks. Dagger symbol indicates mice lost at that time point and therefore subsequent data not collected. S.D. HCV-RNA signal is detected but <3.2 (log IL/ml). N.D. HCV-RNA signal is not detected

Full size image
Table 3 HCV NS3/4A and NS5A RAVs in mice prior to DAA retreatment
Full size table

Finally, we tried to administer telaprevir and GS-558093 for 4 weeks to mice that showed treatment failure by asunaprevir/daclatasvir therapy or ledipasvir/GS-558093 therapy. We confirmed that RAVs against telaprevir in the NS3/4A region, including V36A/M, T54A, R155 K/T/Q and A156S, were not detected (Table 3). Telaprevir/GS-558093 therapy achieved sustained eradication of the mutant virus or the end-of-treatment response (Fig. 2b).

Discussion

The combination of ledipasvir and sofosbuvir is one of the most potent therapeutic regimens for HCV genotype 1 infection and is thus recommended as a first line of therapy in the AASLD guidelines. A monotherapy study of ledipasvir in treatment-naïve genotype 1b HCV-infected subjects conducted early in the clinical phase revealed the emergence of the NS5A Y93H variants after administration [12]. Site-directed NS5A mutants such as L31M or Y93H showed reduced susceptibility to ledipasvir in vitro. [12]. Therefore, such NS5A variants are believed to confer resistance to ledipasvir, like daclatasvir. The ledipasvir/sofosbuvir regimen administered for 12 weeks in treatment-naïve Japanese patients with chronic genotype 1 (1b and 1a) infection resulted in a 100 % SVR rate in patients with baseline NS5A variants [9]. However, the efficacy of ledipasvir/sofosbuvir in treatment-experienced Japanese patients with genotype 1 infection and pre-existing NS5A variants after failure to respond to daclatasvir/asunaprevir has not been established. A combination with asunaprevir and daclatasvir therapy has been applied to more than 40,000 patients in Japan, and approximately 15 % of them most likely experienced viral breakthrough or relapse. Thus, the consideration of re-treatment options for those patients is of great importance. Friborg et al. [13] previously assessed this effect in vitro using replicon elimination assay and demonstrated that the ledipasvir/sofosbuvir regimen still efficiently eliminates the D168V, L31M and Y93H triple variant. However, the in vitro elimination assay only evaluates the decline of the number of cells replicating HCV under drug administration, and whether this really recapitulates in vivo conditions is unsure. In the present study, we assessed this effect in vivo for the first time and demonstrated that the NS3/4A D168V and NS5A L31V and Y93H triple mutant virus appearing in patients with asunaprevir/daclatasvir failure was resistant to the ledipasvir/GS-558093 regimen but still susceptible to a combination of telaprevir and GS-558093.

The present study has some limitations when applying the findings to patients. First, the optimal dose and dosing period of each DAA may be different between humans and mice. However, our therapeutic regimens seem to recapitulate patient settings because the administration of either asunaprevir/daclatasvir or ledipasvir/GS-558093 quickly reduced serum HCV RNA levels and achieved end-of-treatment response (in this case at 4 weeks after starting therapy) and SVR for the wild-type virus except for two mice in the ledipasvir/GS-558093 group. The same protocols were applied for the triple mutant virus. The asunaprevir/daclatasvir regimen was not effective for the mutant virus, as expected. The ledipasvir/GS-558093 regimen reduced the HCV RNA levels but failed to achieve the end-of-treatment response or SVR. These results indicate that the triple mutant virus is relatively resistant to the ledipasvir/GS-558093 regimen compared with the wild-type virus under these experimental conditions.

The second limitation we have to note is the use of GS-558093, not sofosbuvir itself. Since sofosbuvir was still under development when starting this work, we could not obtain sofosbuvir itself for this experiment. GS-558093 is a NS5B polymerase inhibitor that competes with natural nucleotides, thereby causing termination of RNA replication in the nascent viral genome [11]. There may be criticism about whether the GS-558093 used in this experiment has the same effect as sofosbuvir used in the clinic. Indeed, the ledipasvir/GS-558093 regimen only achieved SVR in two of four mice in our experimental conditions. If the ledipasvir/sofosbuvir regimen in the clinic is more potent compared with ledipasvir/GS-558093 in this experiment, there may be a possibility that the ledipasvir/sofosbuvir regimen is still partially effective for patients with asunaprevir/daclatasvir failure.

The present study suggests that, for the triple mutant virus, combination therapies of DAA without cross-resistance are more potent than ledipasvir/sofosbuvir therapy. From this point of view, we may have several options. One is sofosbuvir plus a first-generation, linear-structured protease inhibitor such as telaprevir applied in the present study. Most first-generation, second-wave, macrocyclic protease inhibitors, such as simeprevir, vaniprevir and paritaprevir, are not effective for the D168V variant, similar to asunaprevir, and would not be recommended [14, 15]. Recently, sofosbuvir combined with the non-nucleoside polymerase inhibitor was reported to be effective for genotype 1 infection [16] and might be a choice. Other options may be sofosbuvir plus non-selective antiviral pegylated interferon and ribavirin. This regimen has not been approved in Japan but is used globally [17] and thus may be the most realistic regimen of choice. The prolonged duration of ledipasvir/sofosbuvir or ledipasvir/sofosbuvir added with ribavirin might be another choice.

While RAVs are generally less proliferative than the wild-type virus, the substitution of NS5A Q54H compensates the reduction in the replication capacity of the L31 V-Y93H double mutant HCV [18]. We did not find this substitution in the present samples in either the patient or mice (data not shown). The fact that the present triple mutant variant stably infected the patient in the long-term and showed a similar proliferative capacity as the wild-type virus in mice (Fig. 1) suggests that other mechanisms might have occurred for replication fitness. On the other hand, the frequency of the D168F variant became higher in mice than in the patient (Table 2). The impact of this substitution on viral replication is largely unknown, but it might be adapted to humanized liver chimeric mice.

The NS5B S282T substitution is the primary mutation selected by sofosbuvir in vitro and by site-directed mutagenesis, conferring reduced susceptibility to sofosbuvir [19]. Sofosbuvir is generally considered to possess a high genetic barrier to resistance. Indeed, in an analysis of all patients with treatment failure during phase III trials of sofosbuvir plus ribavirin with or without peginterferon, the S282T substitution was not detected by deep sequencing in any of the 225 patients [20]. In the present study, we did not find the NS5B S282T variant after ledipasvir/GS-558093 therapy or asunaprevir/daclatasvir therapy in mice. This allowed us to choose the GS-558093-containing regimen for the re-treatment option for those mice.

In the present study, we demonstrated that the ledipasvir/NS5B nucleotide polymerase inhibitor GS-558093 regimen is relatively ineffective for RAVs emerging after the treatment failure of asunaprevir/daclatasvir therapy. Cross-resistance should be carefully considered in the era of DAA, especially when re-treating patients who failed to achieve SVR by DAA therapies with a low genetic barrier to resistance.