Telaprevir-based treatment effects on hepatitis C virus in liver and blood

Authors


  • Potential conflict of interest: Dr. Botfield is employed by Vertex. Dr. Chakilam is employed by and owns stock in Vertex. Dr. Dimova received grants from Vertex and Merck. Dr. Jacobson consults for advises, is on the speakers' bureau for and received grants from Bristol-Myers Squibb, Gilead, Roche/Genentech, Merck/Schering-Plough, and Vertex. He consults for advises, and received grants from Abbott, Achillion, Boehringer Ingelheim, Novartis, and Tibotec/Janssen. He consults for and advises Enanta, GlaxoSmithKline, Idenix, and Kadmon. He received grants from Pfizer. Dr. Jiang is employed by and owns stock in Vertex. Dr. Kwong owns stock in Vertex. Dr. Penney is employed by and owns stock in Vertex. Dr. Sawant is employed by and owns stock in Vertex. Dr. Sullivan is employed by and owns stock in Vertex. Dr. Talal advises and received grants from Gilead. He advises Merck, Abbott, Roche, and Tibotec/Janssen. He received grants from Merck/Schering-Plough. He was on the speakers' bureau for Roche/Genentech and Vertex. Dr. Zhang is employed by and owns stock in Vertex.

  • This study was supported by a grant from Vertex Pharmaceuticals Incorporated, which also provided analytic support in the conduct of the study. This study was also supported by grants from the Clinical and Translational Science Center (ULI RR024996) at Weill Cornell and the Greenberg Foundation for Biomedical Research.

Abstract

Understanding hepatitis C virus (HCV) replication has been limited by access to serial samples of liver, the primary site of viral replication. Our understanding of how HCV replicates and develops drug-resistant variants in the liver is limited. We studied 15 patients chronically infected with genotype 1 HCV treated with telaprevir (TVR)/pegylated-interferon alpha/ribavirin. Hepatic fine needle aspiration was performed before treatment and at hour 10, days 4 and 15, and week 8 after initiation of antiviral therapy. We measured viral kinetics, resistance patterns, TVR concentrations, and host transcription profiles. All patients completed all protocol-defined procedures that were generally well tolerated. First-phase HCV decline (baseline/treatment day 4) was significantly slower in liver than in plasma (slope plasma: −0.29; liver, −0.009; P < 0.001), whereas second-phase decline (posttreatment days 4-15) did not differ between the two body compartments (−0.11 and −0.15, respectively; P = 0.1). TVR-resistant variants were detected in plasma, but not in liver (where only wild-type virus was detected). Based upon nonstructural protein 3 sequence analysis, no compartmentalization of viral populations was observed between plasma and liver compartments. Gene expression profiling revealed strong tissue-specific expression signatures. Human intrahepatic TVR concentration, measured for the first time, was lower, compared to plasma, on a gram per milliliter basis. We found moderate heterogeneity between HCV RNA levels from different intrahepatic sites, indicating differences in hepatic microenvironments. Conclusion: These data support an integrated model for HCV replication wherein the host hepatic milieu and innate immunity control the level of viral replication, and the early antiviral response observed in the plasma is predominantly driven by inhibition of hepatic high-level HCV replication sites. (Hepatology 2014;60:1825–1836)

Abbreviations
ALT

alanine aminotransferase

AST

aspartate aminotransferase

CI

confidence interval

CIA

coefficient of individual agreement

CNB

core needle biopsy

FNA

fine needle aspiration

HCV

hepatitis C virus

IC50

half maximal inhibitory concentration

IFN

interferon

IL

interleukin

IQR

interquartile range

ISG

IFN-stimulated gene

NR

nonresponder

NS

nonstructural protein

P/R

pegylated-interferon/ribavirin

SDs

standard deviations

SVR

sustained virological response

T/P/R

telaprevir, pegylated-interferon, and ribavirin

TVR or T

telaprevir

WT

wild-type

Chronic hepatitis C virus (HCV) infection affects ∼120 million people worldwide and nearly 5 million people in the United States.[1, 2] In other viral infections, such as human immunodeficiency virus, a high viral replication rate is often associated with a cytopathic effect, limiting viral replication by killing host cells in combination with a strong immune response, thereby limiting viral replication and spread. In contrast, HCV can persist for decades in a human host, despite a high replication rate of 1012 virions per day.[3] Chronic HCV infection can result in chronic inflammation that promotes ongoing liver damage.[4] Addition of HCV protease inhibitors, such as telaprevir (TVR; T) and boceprevir, to pegylated-interferon/ribavirin (P/R) has been shown to enhance efficacy and may shorten treatment duration.[5, 6] These agents act by blocking polyprotein processing and formation of the HCV replication complex, thereby inhibiting viral replication.

Difficulty in intrahepatic sampling has been a limitation in the understanding of HCV biology. Kinetic studies in patients receiving interferon (IFN) monotherapy or P/R revealed a biphasic decline in plasma HCV RNA levels, wherein the first phase corresponds to clearance of viral particles from circulating blood and the second phase corresponds to the decay of viral replication in infected hepatocytes or hepatocyte loss.[3, 7] More recent studies in patients receiving TVR monotherapy or telaprevir, pegylated-interferon, and ribavirin (T/P/R) described a biphasic decline wherein both the first- and second-phase declines were more rapid than that obtained with P/R.[8, 9] Earlier studies during P/R treatment suggested that baseline intrahepatic IFN-stimulated gene (ISG) expression levels are increased and that the degree of induction is predictive of treatment efficacy, suggesting that liver transcript analysis may identify predictors of response.[10, 11]

Although a few studies have employed core needle biopsy (CNB) for liver sampling in HCV infection,[11, 12] serial measurements using CNB are difficult. An alternative is liver fine needle aspiration (FNA), which has significantly reduced morbidity, compared to standard liver CNB.[13] We performed serial liver FNAs at baseline and during antiviral therapy on 15 patients with genotype 1 chronic HCV infection, 9 treatment-naïve, and 6 P/R previous nonresponders (NRs) to understand the correlates of intrahepatic and peripheral viral kinetics. Specifically, we assessed (1) kinetics of wild-type (WT) and TVR-resistant variants, (2) pharmacokinetics of TVR, and (3) host gene expression. Whereas the mechanisms of action and efficacy of the treatment regimens are well understood, within the liver, in vivo exposure to TVR, degree of inhibition of viral replication, and rate of clearance of HCV RNA from infected cells are unknown.

Patients and Methods

Fifteen genotype 1 chronic HCV-infected patients, 18-65 years of age, received TVR 750 mg every 8 hours (q8h), pegylated-interferon-alpha-2a 180 μg weekly, and weight-based ribavirin (1,000 or 1,200 mg/day) (T/P/R) for 12 weeks, followed by at least 12 additional weeks of P/R. Nine patients were treatment-naïve (<4 weeks of previous P/R treatment), and 6 were previous NRs to P/R. All patients had stage 1-3 fibrosis, with 11 with stage less than or equal to 2. Written informed consent was obtained from each patient, and the study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki, as reflected in a priori approval by the Weill-Cornell Institutional Review Board (New York, NY).

Patients were hospitalized overnight for treatment initiation. Liver FNAs were performed at baseline (within 7 days before treatment initiation), at 10 hours after initiation of triple therapy, on days 4 and 15, and after a protocol amendment, at week 8 (Supporting Fig. 1).

Plasma HCV RNA levels were measured using the COBAS AmpliPrep/COBAS TaqMan HCV Test (version 1 test with AmpliPrep extraction; Roche Diagnostics, Indianapolis, IN) with a quantification limit of 43 IU/mL and a detection limit of 7 IU/mL for genotype 1.

Statistical Analysis

Statistical analysis was performed using SAS (SAS Institute Inc., Cary, NC) and R software (http://www.r-project.org/). Continuous variables are presented by their means or medians and their standard deviations (SDs) or interquartile ranges (IQRs). Categorical variables are summarized using counts and percentages. HCV RNA viral kinetics in plasma and liver were modeled through nonlinear mixed-effects models adjusted for left censoring of HCV RNA levels. TVR concentrations in plasma and liver over time were compared through a mixed-effects model. Repeatability of the FNA procedure for assessment of intrahepatic HCV RNA levels was evaluated through the coefficient of individual agreement (CIA). Comparison between continuous variables, if not explicitly stated, was conducted through the Wilcoxon signed-rank or rank-sum tests, and between categorical variables, through Fisher's exact test. More detailed information on statistical methods can be found in the Supporting Information.

The following methods are described in the Supporting Information: (1) RNA isolation from liver and determination of HCV RNA copy number; (2) HCV population and clonal sequencing in liver and plasma; (3) gene expression assessment using the nCounter Analysis System (NanoString Technologies, Seattle, WA); (4) TVR measurements; and (5) statistical methods.

Results

Patients

A total of 15 patients received at least one dose of T/P/R. Eleven (73%) patients were Caucasian, 3 (20%) African American, and 1 (7%) Hispanic. Median age was 55 years (IQR, 46.5, 58) and 9 (60%) were male (Supporting Table 2). Eleven patients had HCV genotype 1a, 2 were interleukin (IL) 28B genotype CC, and 2 were genotype TT. All patients completed all protocol-defined FNAs, and the procedure was generally well tolerated; no pain was reported by 8 of 15 patients during or after any of the FNA procedures (Supporting Information).

Treatment Course and Outcome

A total of 9 (60%) patients achieved a sustained virological response (SVR) and 6 did not achieve SVR after T/P/R. Among the latter patients, 4 experienced viral breakthrough, 1 was lost to follow-up at week 8, and 1 self-discontinued on day 15 because of side effects. Among treatment-naïve patients, 7 of 9 (77.8%) achieved SVR, whereas among previous NRs to P/R, 2 of 6 achieved SVR. Among genotype 1a patients, 7 of 11 achieved SVR, and among 1b patients, 2 of 4 achieved SVR. Both IL28B CC genotype patients achieved SVR (7 of 13 for non-CC).

Drug Concentration

Overall TVR concentrations differed significantly in plasma, compared to liver (P = 0.007; Fig. 1A), with median (IQR) liver/plasma ratios of 0.72 (0.4, 1.01), 0.47 (0.33, 0.78), 0.57 (0.34, 0.76), and 0.61 (0.42, 0.84) at each of the four time points when concentrations were assessed. TVR concentrations in human plasma and liver are 5.5- to 14.4-fold and 2.3- to 14.9-fold higher, respectively, than the half maximal inhibitory concentration (IC50) in the HCV replicon[14] and 0.7- to 6.3-fold and 2.3- to 6.1-fold, respectively, above the 90% IC (IC90).[15]

Figure 1.

(A) Plot of the mean (SD) ln-transformed (A) Telaprevir (TVR) concentration in plasma and liver at each time point. (B) Plot of the mean (SDs) log10-transformed HCV RNA levels over time. Undetectable levels in liver and plasma are depicted on the lower limit of detection line (LLOD). The vertical dashed line corresponds to TVR, pegylated-interferon, ribavirin (T/P/R) treatment initiation. Predose corresponds to baseline before treatment initiation. H6, hour 6 after treatment initiation; H10, hour 10; D4, day 4; D8, day 8; D15, day 15; W4, week 4; W8, week 8; W12, week 12. At each time point, the extensions depict the respective SDs.

HCV RNA Decline in Plasma and Liver

In plasma, mean log10-transformed copies/mL HCV RNA levels were 6.5 ± 0.8 at predose, 4.9 ± 0.9 at 10 hours, and 3.2 ± 0.6 on treatment day 4 (Fig. 1B and Supporting Fig. 2). In liver, HCV RNA levels were quantified relative to the total mass of RNA, and mean log10-transformed copies/ng total HCV RNA levels were 4.2 ± 1.2 at predose, 4.0 ± 1.3 at 10 hours, and 3.9 ± 1.2 at 4 days after treatment initiation.

Between baseline and day 4, we observed a slower decrease in hepatic HCV RNA relative to plasma (median log10-transformed HCV RNA decline slope: plasma, −0.29; liver, −0.009; P < 0.001). Between treatment days 4 and 15, median rates of HCV RNA decline in liver and plasma were not significantly different (−0.11 and −0.15, respectively; P = 0.104). All posttreatment week 8 intrahepatic HCV RNA levels were undetectable, and a higher slope of HCV RNA decrease was observed between day 15 and week 8 in liver (−0.04) than in plasma (−0.01; P = 0.02), potentially attributable to low plasma HCV RNA levels.

Modeling HCV RNA in Plasma and Liver

A detailed description of the statistical modeling is given in the Supporting Information. In plasma, HCV RNA levels were modeled according to Neumann et al.,[3] assuming constant effectiveness, accounting for left censoring and adjusting the infected hepatocyte loss rate for patients' sex, treatment-naïve status, and baseline alanine aminotransferase (ALT) levels (Table 1; Supporting Fig. 3). Infected hepatocyte loss rate (δ), which characterizes the second phase of viral decline,[3] was positively correlated with the baseline ALT level, male gender, and treatment-naïve status. Addition of the parameter δ as determined from the liver model (below) did not improve plasma model fit.

Table 1. Results From the Model Fit for HCV RNA in Plasmaa
Parameter Estimates
ParameterEstimateStandard ErrorPr > |t|LowerUpper
  1. a

    Assuming constant effectiveness and adjusting the infected hepatocyte loss rate, δ, for patients' sex (βMale), treatment-naïve status (βTr.naive), and ALT levels (βALT), as determined at screening. Parameter εp is the efficacy of treatment in blocking virion production, c is the rate per virion at which virions are cleared, and t0 is the time at which the drug becomes effective.

εp0.990.002<0.0010.9890.998
c16.503.27<0.0019.4423.57
δ00.120.080.12−0.0380.286
t00.180.02<0.0010.130.22
βALT0.0040.0010.020.00070.0065
βTr.naive0.160.070.04940.00050.32
βMale0.310.07<0.0010.170.45

The plasma model did not fit the viral kinetics observed in the liver well (based on the Akaike Information Criterion corrected for small sample sizes (AICc); Supporting Table 5). As such, the model was modified by setting the virion clearance rate equal to the infected hepatocyte loss rate and setting the treatment efficacy for reducing new infections to 1. Both baseline ALT and aspartate aminotransferase (AST) levels were positively correlated with viral decline (ρ = 0.69, P = 0.006 and ρ = 0.70, P = 0.008, respectively). We observed a high correlation (ρ = 0.93; 95% confidence interval [CI]: 0.78, 0.98) between ALT and AST levels. Model parameters that corresponded to the best fit are shown in Table 1B (also Supporting Fig. 3). The association between viral decline and ALT tended to differ between patients depending upon whether they were infected with HCV subtypes 1a or 1b (Table 1B). No association was observed between inflammation (portal and lobular) and viral decline. We estimated individual values for δ for each patient, as determined by plasma and liver models, and median values (0.58, IQR [0.35, 0.80] vs. 0.44, IQR [0.40, 0.57]) differed significantly between compartments (p = 0.03; Table 1C).

Assessment of Intrahepatic HCV RNA Variability

We assessed the repeatability of the FNA sampling technique based on 33 pairs of liver HCV RNA samples from 12 of the 15 study patients (Fig. 2A). None of the patients had more than one pair of liver HCV RNA samples per time point. In 27% (5 of 18) of pairs with detectable values, we observed a ratio of the larger value to the lower of >3. Twelve pairs had undetectable HCV RNA in one of the FNA passes; of them, five had an HCV RNA value <190 copies/ng. The paired differences in log10 (HCV RNA) measurements are plotted versus their average (Fig. 2B). In general, we observed moderate agreement (based on the CIA) between HCV RNA measurements in liver obtained from the two separate FNA passes (Supporting Information, including Supporting Table 6), with CIA ranging between 0.41 and 0.59 at the different time points. The 95% CI for the ratio between the two samples was (0.003; 906.8-fold).

Figure 2.

(A) Plot of 33 log10-transformed HCV RNA paired samples from liver obtained through two FNA passes on 12 patients. Dashed lines refer to values below the lower limit of detection (LLOD), imputed as half the level of detection. (B) Bland and Altman plot of the paired differences in the two liver log10-transformed HCV RNA samples versus their average. Horizontal solid line corresponds to the mean difference, whereas horizontal dashed lines correspond to the lower and upper 95% limits of agreement (LoA). The plotted symbol is the patient's identification number.

Viral Resistance to TVR

We sequenced WT and TVR-resistant variants in all samples with quantifiable HCV RNA using population sequence analysis and, in a subset of samples (n = 7), by clonal analysis (Table 2). Sequencing was successful in 100% (15 of 15) of baseline plasma and in 80% (12 of 15) of baseline liver samples. Plasma subtype distribution, as determined by the nonstructural protein (NS)3 sequencing method, was 73% (n = 11) 1a and 27% (n = 4) 1b.

Table 2. Results From the Model Fit for HCV RNA in Livera
Parameter Estimates
ParameterEstimateStandard ErrorPr > |t|LowerUpper
  1. a

    Assuming model 3, (as indicated in Supporting Table 5) and adjusting the loss rate δ for patients' ln(ALT) level measured at screening (βALT), subgenotype 1a (β1A), and the interaction between ln(ALT) and subgenotype 1a (β1A*ALT). Parameter P is the rate per infected hepatocyte at which virions are produced.

δ0−6.783.070.046−13.41−0.15
P12.4010.070.24−9.3634.16
βALT2.040.890.040.123.97
β1A6.363.100.06−0.3313.05
β1A*ALT−1.810.890.06−3.740.12
Table 3. Estimated Values of Parameter δ for Each Patient From Plasma and From Liver HCV RNA
Patient ID (PID) no.δ From PlasmaStandard Error of δ From Plasmaδ From LiverStandard Error of δ From Liver
  1. For patient with PID 1007, δ from liver was not estimated because all HCV RNA liver measurements were undetectable. δ is the rate per cell at which infected hepatocytes are lost.

10010.800.100.570.11
10020.910.140.600.10
10030.500.060.340.09
10040.330.060.440.11
10060.300.060.420.08
10070.370.12  
10080.880.170.780.14
10090.590.090.430.09
10100.350.060.440.08
10110.950.122.611.03
10120.840.100.590.09
10130.690.100.400.08
10140.580.080.540.10
10150.290.060.270.09
10160.740.100.270.10
Table 4. Population Sequence (NS3 Region) of Paired Plasma and Liver Samplesa
 Patient ID no.Genotype/SubtypeTime PointPlasmaLiver
HCV RNA (IU/mL)TVR-Resistant Variants ObservedPrimary Liver SampleSecond Liver Sample
HCV RNA (Copies/Total RNA)TVR-Resistant Variants ObservedHCV RNA (Copies/Total RNA)TVR-Resistant Variants Observed
  1. Abbreviations: D-7, baseline (before dose); D1 6H, day 1 six hours post dose; D1 10H, day 1 ten hours post dose; D4, day 4; ND, not done; NA, not available; LLOQ, lower limit of quantitation; IU, international units.

  2. a

    Five discordant pairs and 23 concordant pairs.

Discordant pairs10061aD41,109T54A, A156T, R155K33,308WT<LLOQNA
10101aD41,374V36A, R155K/G3,371WT<LLOQNA
10111bD1 10H141,519T54T/A2,466WTNDND
10141aD42,018V36V/M92,901WT53,842WT
10151bD43,662A156A/T73,763WTNDND
Concordant pairs10021aD-73,050,727WT18,025WT24,118WT
D1, 6H60,054WTndNANAWT
10031aD-74,081,131WT18,427WT18,069WT
D1 10H549,890WT17,204WT29,213WT
10041aD-73,878,672WT30,271WT18,169WT
D1 10H172,565WT75,929WTNDND
D45,111WT75,336WT60,953WT
10061aD-71,365,871WT63,457WT<LLOQNA
10081aD-7645,930WT91,886WT<LLOQNA
D1 10H6,112WT19,028WTNDND
10091aD-7246,830WT46,950WT152,915WT
D1 10H11,445WT23,875WTNDND
10101aD-71,323,535WT41,059WTNDND
10111bD-72,515,642WT494WTNDND
10121aD-71,530,810WT47,516WTNDND
D1 10H29,674WT72,019NA21,895WT
10131aD-7540,776WT174,345WT77,401WT
D1 10H112,185WT492,869WT73,089WT
10141aD-72,937,704WT83,466WT77,882WT
D1 10H45,201WT8,142WT106,888WT
10151bD-72,518,054WT114,019WTNDND
D1 10H39,943WT106,458WTNDND
10161bD1 10H23,407WT55,800WTNDND

Before treatment, only WT virus was detected in both plasma and liver in all patients. In contrast, during or after the treatment phase, TVR-resistant variants[16, 17] were observed in 7 patients (2 SVRs and 5 non-SVRs) in plasma (V36M/A, T54A/S, R155K/G, A156T/V, R155M+A156A/T, and V36M+R155K; Supporting Table 7; Fig. 3), whereas only WT virus was detected in liver, even for those visits at which resistant variants were observed in plasma (Table 4 and Supporting Table 8).

Figure 3.

Population sequence of plasma (NS3/4A) and liver FNA (NS3) of patients who did not obtain an SVR. See figure legend for symbol explanation. LLOD, lower limit of detection; LLOQ, lower limit of quantitation.

Phylogenetics of Liver and Plasma Sampled Virus

Liver and plasma viral populations were compared for 12 baseline samples and 16 on-treatment visits using population sequence analysis. Highly sensitive clonal sequence analysis, able to detect minor viral populations (≥5%), was performed on samples from 3 patients. A total of 14 isolates (median 73 [range, 63-89] and 77 [range, 59-94] clones) were sequenced from each plasma and liver FNA isolate, respectively. These patients were selected because they had the largest number of genomic differences between liver and plasma populations based on population sequence results within the NS3/4A proteins.

The phylogeny generated from the population sequence analysis suggested no compartmentalization between liver and plasma. All isolates from each patient clustered together to the exclusion of any isolates from another patient, regardless of the timing of sampling (minimum bootstrap support: 93%; Fig. 4). Significantly, this result was supported by clonal analysis, indicating lack of clustering between liver and plasma isolates from 3 patients sampled at two or more time points (Fig. 5). These data suggest that both liver and plasma isolates sampled the same underlying viral populations.

Figure 4.

Unrooted phylogenetic trees of population sequence of the NS3 region of plasma (black) and liver FNA (red) isolates, for both genotype (GT) 1a (A) and 1b (B). The sample identification number (ID) of each patient is indicated, as is the time at which the sample was collected (D-7, 1 week before treatment initiation; D1, 10 hours after the first treatment; D4, day 4). To assess method variability, a subset of isolates was processed in replicate, with the parenthetical number after the sample ID. Monophyly of virus from each patient is well supported (minimum bootstrap support of 93%). In no instance did the tree support monophyly of virus from either compartment.

Distance scale: scale that represents the number of differences between sequences (e.g., 0.1 means 10% differences between two sequences).

Figure 5.

Phylogenetic analysis of clonal sequence data. Clones derived from liver and plasma isolates are identified by filled triangles and squares, respectively. Where multiple amplifications were performed (to avoid amplification artifacts), they are indicated with the color code. For reference, the population sequence analysis of each isolate was also included in the phylogeny and is depicted in the open symbols. Each panel is drawn to its own scale, with the scale bar indicating the number of substitutions per nucleotide position.

Gene Expression Between Compartments

As expected, we observed highly significant differences between liver and blood samples. At baseline, samples could be differentiated based on tissue type, rather than donor origin (i.e., blood and liver samples from a patient never clustered together, but rather clustered with the samples of the same tissue; Fig. 6A). Differentiation between tissue types was observed in 203 of 267 loci assessed (P < 0.05), with 154 of these loci having significantly different expression levels after Bonferroni's correction (Supporting Table 7; P value range: 1.2e-17, 1.6e-4).

Figure 6.

(A) Robust baseline differential expression between tissue types. The heatmap displays the intensity of the signal of each gene (arranged in columns) as indicated in the color legend along a relative scale, with each sample represented by a single row. (B) Eight loci previously implicated in the literature (see text) as being associated with responsiveness during pegylated interferon/ribavirin treatment tended to show differences between SVRs and those who did not obtain an SVR (NRs).

We also assessed degree of induction of these loci by T/P/R on day 1, relative to baseline in blood and liver, between SVR and non-SVR patients. After type 1 error correction, no loci achieved statistical significance for prediction of treatment outcome. However, we observed a trend toward statistical significance between SVR and non-SVR patients for intrahepatic induction levels of 35 loci (P < 0.05). Furthermore, 8 of 12 loci previously identified as predictive of P/R-based treatment outcome[18, 19] were significantly higher in SVR patients (P < 0.05 before type 1 error correction; Fig. 6B).

Discussion

In this study, we evaluated four distinct aspects of HCV infection and treatment (viral decay, drug exposure, HCV genomic variation, and transcriptional profiling) to compare liver and blood before and after initiation of antiviral therapy. We observed the following: (1) The initial decline in intrahepatic HCV RNA lagged, compared to plasma, whereas the second-phase decline was similar; (2) TVR-resistant variants were detected in plasma, but not in liver (where only WT virus was detected); (3) No compartmentalization in liver and plasma viral populations was observed, based upon NS3 sequence analysis; (4) As expected, gene expression profiling revealed strong tissue-specific expression signatures in baseline and on-treatment samples. However, loci previously described as associated with SVR in P/R-treated patients were not predictive of treatment outcome in T/P/R-treated patients in this study with a small sample size that included a large subset of previous P/R NRs; (5) Human intrahepatic TVR concentration, measured for the first time, was lower, compared to plasma, on a mass per volume basis.

The relationship between HCV RNA decline in human liver and blood has not been previously assessed largely because of the difficulty of serially sampling the liver using conventional CNB. In plasma, our modeling results (virion clearance rate c = 16.5 and mean infected hepatocyte loss rate δ = 0.61) were comparable to other studies (c = 13.4 and δ = 0.589; c = 15.93 and δ = 0.29-0.628) in T/P/R-treated patients. Unlike earlier studies, we adjusted our model for left censoring in the data. We performed repeated FNA sampling in liver, and in comparison to plasma, we observed a significantly slower decrease in HCV RNA during the first phase of viral decline. Whereas the results we obtained regarding virus decline in plasma during the first phase were consistent with previously published reports, in liver there was very little change in HCV RNA during this time interval. Previously published models of viral decline in plasma[3, 9] fit our plasma data well, but not the liver data. In order to optimize the fit of the liver model, we set the viral clearance rate (c) equal to the loss rate of productively infected cells (δ) and the efficacy of treatment in reducing new infections equal to 1. Consequently, as indicated by the significantly slower rate of intrahepatic viral decline, when compared to blood expressed by the loss of HCV cellular replication and/or the death rate of infected cells, δ, we conclude that HCV RNA decay is markedly delayed in liver, compared to plasma.

To compare the hepatic and peripheral viral populations, we performed population and clonal sequence analysis using the NS3 region because its gene product, the HCV protease, is the TVR target. Although this protein is highly conserved, relative to many other HCV proteins, NS3 sequence analysis was sufficiently sensitive to demonstrate separate phylogenies for each patient's viral quasi-species. Overall, we did not observe any substantial differences in NS3 hepatic or peripheral viral population sequences sampled from a single patient at the same time point. Interestingly, we observed TVR-resistant variants exclusively in blood during the first phase of viral decline. These variants have been shown to be “uncovered” when replication of the much larger population of WT virus is inhibited[16, 20]; their presence per se was not associated with treatment failure. Alternatively, because NS3 inhibition has recently been shown to interfere with a multitude of viral processes, including assembly and release,[21] replication complexes that contain protease resistance mutations might preferentially be exported to the plasma compartment.

These findings are consistent with TVR's primary role in inhibition of WT and lower-level TVR-resistance variants (defined as a 3- to 25-fold increase in IC50, compared to WT) conferred by V36A/M, T54A/S, R155K/T, and A156S HCV protease variants.[16] In patients who did not achieve an SVR with treatment, the resistant population became predominant in plasma. Because mutation is a clonal process and only a very limited sample of liver is evaluated with FNA, it is not surprising that virus detected in liver is WT, at the same time that the product of the mutant clones is detected in plasma.

Taken together, these results might be explained by at least two models: (1) delayed IFN clearance or (2) high- and low-level sites of HCV replication. In model 1, between baseline and posttreatment day 4, TVR inhibits replication of the majority of WT and low-level TVR-resistant variants, but IFN-induced clearance of infected cells has not yet commenced. TVR-induced inhibition of replication of these variants is permissive for replication of primarily higher-level TVR-resistant variants, which were observed on day 4. The absence of resistant variants in liver at the same time point suggests that these resistant variants may be a minority population of the total infection, compared to the majority of nonreplicating WT virus that has not yet been cleared by the host's immune response. Alternatively, they may be absent within the microenvironment sampled. In summary, HCV RNA concentration and sequencing data together support a model in which TVR rapidly and efficiently inhibits viral replication in hepatocytes, but host immunity is required to clear infected cells from the liver, as observed during the second phase of viral decline during P/R treatment.

In model 2, we posit that intrahepatic HCV replication is not uniform, that is, a few sites of high-level replication are surrounded by many sites of low-level replication. In vitro studies have described limited spread of infection, ∼0.5% of Huh7 cells with HCV Japanese fulminant hepatitis type 1 virus infection.[22] In vivo studies have illustrated foci of HCV-infected cells in liver samples from patients with chronic HCV infection.[23] Furthermore, other studies[24] showed that HCV genomic RNA was variably distributed throughout HCV-infected livers and that HCV replicative-intermediate RNA is distributed in a focal pattern. Thus, before treatment, the majority of plasma HCV RNA may have originated from high-level replication sites. Because the NS3 sequence from plasma and liver shows no compartmentalization, the host is likely to be the major determinant of the level of viral replication. HCV replication has been shown to be higher in cultured cells where part of the innate immune response is knocked out.[25] Additionally, hepatocyte production of CXC chemokine receptor 3–associated chemokines, the primary promoter of intrahepatic inflammation, is distributed focally throughout the hepatic parenchyma.[26] With initiation of antiviral therapy, we hypothesize that inhibition of WT virus in the high-level replication sites contributes to the bulk of the first-phase decline. With treatment continuation, the lower-level replication sites will decline as a result of prolonged T/P/R suppression in combination with host immunity. Resistant variants are created through the error-prone HCV replication machinery and are present at a higher level in foci with high replication levels. Furthermore, these sites are the primary contribution to the plasma quasi-species from which resistant variants arise and are uncovered when WT and low-level resistant variants are inhibited by T/P/R.

Hepatic TVR concentrations were lower, relative to plasma, and significantly different than in preclinical studies in rats or dogs where liver concentrations were 10- to 20- or 1- to 3-fold higher, respectively, relative to plasma.[14, 27] Although liver/plasma ratios for some protease inhibitors measured in vivo in preclinical species have been shown to have a reasonable correlation to the human situation,[27] our data suggest that TVR preclinical studies in animals may not entirely reflect the in vivo human situation. TVR has not been found to be a substrate for uptake transporters,[17] although it is a substrate for, and an inhibitor of, P-glycoprotein.[28]

In contrast to earlier studies performed in P/R-treated patients,[11, 12, 18, 19] we did not identify an ISG signature in plasma or blood that was predictive of treatment outcome. These findings may result from differences in therapeutic regimen (P/R vs. T/P/R) or patient characteristics. Additionally, the degree of hepatic ISG induction under P/R treatment may be a more significant driver of SVR probability than it is in a treatment regimen that includes a direct-acting antiviral.

To test the hypothesis that the liver is uniformly infected with HCV and that different microenvironments mount comparable immune responses during infection, we compared HCV RNA and host transcriptional profiling across multiple FNA isolates in the same patient. Overall, HCV RNA levels were moderately consistent between two FNA isolates sampled at a single time point: for the majority of FNA samples, HCV RNA measurements were highly correlated. Measurement variability could have resulted from differential cell-type composition within the sample, different stages of fibrosis between samples, or differences in the microenvironment surrounding the area sampled that could have resulted in differences in the infected cell rate or in the intrinsic viral replication rate within that microenvironment. Interestingly, we also noted intrahepatic variability in gene expression profiles, as indicated by poor agreement between loci measured on two separate FNA samples. Whereas this analysis was retrospective and different analytical methods were used to measure each isolate, it crudely compared transcriptional levels across isolates; the results suggest variability in expression of loci within the liver. Taken together, these results suggest existence of marked differences within liver microenvironments in terms of the quantity of HCV present and host transcriptional activity.

In this study, we employed the FNA technique in a novel manner (serial hepatic sampling over a short interval during antiviral treatment). Though potential limitations of FNA sampling include variable quantities of liver and cross-contamination by plasma, expression signatures on liver samples and the vast majority of sampled loci were differentially expressed between samples from the two compartments. Small sample volumes obtained by FNA may also have affected our ability to detect resistant variants in the liver. Because of the difficulty obtaining samples by aspiration, patients with cirrhosis were excluded, although advanced fibrosis could have affected the HCV RNA decline rates or intrahepatic TVR concentration. These results suggest that liver comprised the major portion of the FNA sample, with minimal plasma contamination. Additionally, FNA TVR concentrations in the liver were lower than in plasma, supporting minimal cross-contamination. Overall, our results support an integrated model for HCV replication where: (1) HCV in plasma is the progeny of HCV replicating in the host liver; (2) Different hepatic microenvironments support a few high-level and a larger number of low-level sites of viral replication; (3) When patients are treated with a potent antiviral drug, the first-phase rapid viral decline observed in plasma is predominantly driven by the inhibition of WT HCV at high-level sites of viral replication in the liver; (4) Drug-resistant variants, generated by the error-prone HCV replication site, are also more likely to originate from sites of high-level replication. For anti-HCV therapy to be efficacious, it must inhibit the virus at all sites of viral replication in the liver. The first-phase decline of plasma HCV precedes clearance of virus in the liver. The second slope of virus decline in the plasma parallels that in the liver and is a better predictor of overall viral clearance rate.

Acknowledgment

The authors acknowledge the following individuals for their assistance with the conduct of the study: Ginevra Castagna, R.N., Beatrice Mesidor, and Gertrudis Soto of Weill-Cornell Medical College; the staff of Weill-Cornell Medical College Clinical and Translational Science Center; Lindsay McNair, M.D., David Alexandrian, M.D., and Scott McCallister, M.D. (each former employees of Vertex Pharmaceuticals Incorporated). In addition, the authors acknowledge the following individuals for assistance with the laboratory analyses of the biological samples from the FNA: Yaroslava Makeyeva and Samantha Benjamin of Weill-Cornell Medical College; Benjamin Shames, Andrew Jayaraj, Ph.D., and John Alam, M.D. (each a former employee of Vertex Pharmaceuticals Incorporated); and Joan Spanks, Andrew Davis, Ann Tigges, Doug Bartels, Ph.D., Katherine McDermott, Kevin Kelliher, Catherine Phillips, and Tara Kieffer, Ph.D., of Vertex Pharmaceuticals Incorporated. A.H.T. served as sponsor-investigator for the study and had full access to the data. A.H.T. had final responsibility for the decision to submit the manuscript for publication.

Ancillary