SEARCH

SEARCH BY CITATION

Dixit NM, Layden-Almer JE, Layden TJ, Perelson AS. Modelling how ribavirin improves interferon response rates in hepatitis C virus infection. Nature 2004; 432: 922924. (Reprinted with permission from Nature Publishing Group, http://www.nature.com.)

Abstract

  1. Top of page
  2. Abstract
  3. Comments
  4. References

Nearly 200 million individuals worldwide are currently infected with hepatitis C virus (HCV). Combination therapy with pegylated interferon and ribavirin, the latest treatment for HCV infection, elicits long-term responses in only about 50% of patients treated. No effective alternative treatments exist for nonresponders. Consequently, significant efforts are continuing to maximize response to combination therapy. However, rational therapy optimization is precluded by the poor understanding of the mechanism(s) of ribavirin action against HCV. Ribavirin alone induces either a transient early decline or no decrease in HCV viral load, but in combination with interferon it significantly improves long-term response rates. Here we present a model of HCV dynamics in which, on the basis of growing evidence, we assume that ribavirin decreases HCV infectivity in an infected individual in a dose-dependent manner. The model quantitatively predicts long-term response rates to interferon monotherapy and combination therapy, fits observed patterns of HCV RNA decline in patients undergoing therapy, reconciles conflicting observations of the influence of ribavirin on HCV RNA decline, provides key insights into the mechanism of ribavirin action against HCV, and establishes a framework for rational therapy optimization.

Comments

  1. Top of page
  2. Abstract
  3. Comments
  4. References

The current standard of care for chronic hepatitis C is a combination of pegylated interferon-α (Peg-IFN-α) and ribavirin (RBV). More than 50% of treated patients achieve sustained viral response (SVR), reflecting the remarkable progress made clinically throughout the past 15 years.1, 2 However, the mechanisms of action for both interferon-α (IFN-α) and RBV remain poorly understood.

IFN-α was first approved as monotherapy for treatment of chronic hepatitis C in 1990. The efficacy of IFN-α correlates with the duration of treatment, and the dose and continuity of drug exposure. Viral response to IFN-α is rapid but incomplete, resulting in a biphasic decline in serum hepatitis C virus (HCV) RNA levels.3 Using mathematical modeling and various data-fitting techniques, Neumann et al. concluded that the majority of IFN-α antiviral activity resulted from inhibiting viral production or release from infected hepatocytes.3 This was further supported by in vitro studies showing that Huh-7 cells harboring subgenomic HCV replicons could be “cured” by adding IFN-α to the cell cultures.4, 5 This clearly demonstrated that IFN-α could effectively inhibit viral replication and eliminate intracellular viral RNA through a noncytolytic mechanism. IFN-α is known to exhibit pleiotropic antiviral activities through several well-characterized effector molecules, including an enzyme known as double-strand RNA-specific adenosine deaminase (ADAR1), which can induce A-to-G mutations in viral genomes.6 Therefore, IFN-α may introduce detrimental mutations into HCV genomes.7

One major drawback of IFN-α monotherapy is the high relapse rate. Many patients who achieve end-of-treatment response (ETR) with undetectable serum HCV RNA become viremic again during follow-up,8 suggesting an incomplete viral clearance or control. Little is known about the cause of this relapse. RBV, a small nucleoside inhibitor, was found to uniquely enhance ETR and prevent relapse, thereby increasing the rate of SVR.9 A recent study by Pawlotsky et al. re-examined the antiviral effect of RBV with more frequent sampling during the early phase of RBV monotherapy.10 They convincingly demonstrated that RBV was able to induce a weak, transient, yet significant antiviral response, suggesting that RBV does not simply serve as an adjunct therapy to IFN-α. Somehow, RBV can synergistically enhance the IFN-α action.

Like IFN-α, the mechanisms of action of RBV are pleiotropic and poorly characterized. Two new theories have been proposed to explain the effect of RBV in HCV therapy: (1) immunomodulation toward T helper 1 responses in favor of antiviral immunity and (2) lethal mutagenesis to the HCV RNA genome.11, 12 Immunomodulation may result in cytolytic “killing” as well as noncytolytic “curing” of infected cells (Fig. 1). Lethal mutagenesis can reduce viral infectivity and fitness by increasing mutations in the genome beyond a threshold that “error catastrophe” ensues.12

thumbnail image

Figure 1. HCV dynamics and mechanisms of action of interferon-α (IFN) and ribavirin (RBV). Productively infected hepatocytes produce and release both infectious virions (●) and noninfectious virions (○) at a rate of p. IFN inhibits HCV production by a factor of (1-ϵ) and ϵ represents IFN effectiveness. RBV introduces mutations into the HCV genome and increases the production of non-infectious virions by a factor of ρ. This will in turn reduce the infectivity measured at the next round of infection by a factor of (1-ρ). The fraction of noninfectious virions induced by RBV is represented by ρ. Both IFN and RBV are proposed to turn over infected cells by immune killing (k) through T lymphocytes or natural killer (NK) cells or noncytolytic “curing” (q) through cytokines. Hence, δ (loss rate on infected cells) = k (killing rate) + q (curing rate). Target cells (uninfected hepatocytes) can be infected productively by the infectious virions, but not by noninfectious virions. Target cells can also originate from infected cells “cured” by IFN and RBV. Inhibition is represented by a perpendicular sign (⟂), whereas an arrow with a plus sign indicates enhancement.

Download figure to PowerPoint

Proving these theories to be clinically relevant has been difficult. The immunomodulatory activity of RBV can be overshadowed by IFN-α therapy. RBV mutagenesis has been demonstrated using poliovirus as a surrogate that allows the functional selection of a mutation against a biomarker.13 Demonstration of RBV-induced mutations in HCV has been controversial because of the lack of functional assays. Large-scale sequencing following error-prone reverse-transcriptase polymerase chain reaction amplification is the only choice. The ability to detect RBV-increased mutations has been investigator- or system-dependent with a range of findings, from equivocal to statistically significant.7, 10 The highly dynamic nature of HCV quasispecies adds complexity to the measurement of slightly increased mutation rate based on direct sequencing of a small number of clones. There is a need to devise new methodology to study the mutagenesis theory.

Mathematical models have been used to delineate mechanisms of action of antivirals. Herrmann and colleagues, in a study to compare viral decay kinetics between pegylated IFN-α alone and pegylated IFN-α or IFN-α plus RBV, were among the first to use mathematical models to study the mechanisms of action of RBV.14 IFN-α effectiveness (ϵ) measured by the log drop in viral titers during the first day of treatment, was shown to be lower at dosages approved by the U.S. Food and Drug Administration (ϵ = 0.36-0.67). A unique triphasic viral decline was observed in a proportion of patients. The third phase decline was interpreted as representing an accelerated loss rate of infected cells by RBV and correlated with improved SVR. The acceleration of loss rate of infected cells (δ) was modeled by an inflation factor M and attributed to an enhanced immune killing, indicating immunomodulation as the mechanism of action of RBV (Fig. 1).14 However, this enhanced immune killing was not supported by the more pronounced alanine aminotransferase/aspartate aminotransferase suppression in the RBV groups. The authors attributed this to the noncytolytic “curing” of infected cells devoid of significant liver injury (Fig. 1). Interestingly, it was shown previously that a small and transient rise in alanine aminotransferase in early phases of IFN-α therapy correlated with favorable treatment responses.15 Nevertheless, Herrmann et al. were unable to rule out the possibility of RBV mutagenesis contributing to the accelerated loss of infected cells. The effect of RBV on de novo infection (β), more specifically on the noninfectious virion compartment (η, fraction of noninfectious virions), was not analyzed. Perhaps the model was unable to distinguish between two variables simultaneously (i.e., effect on loss rate of infected cells versus effect on infectivity).

In a seminal report published in Nature last December, Dixit and colleagues further investigated the mechanism of action of RBV using mathematical models.16 High-dose (10 MIU) IFN-α was administered daily to maximize its effectiveness. Under the condition of high IFN-α effectivenss (ϵ > 0.95), viral titer changes over the first 28 days of therapy could be fitted to a biphasic decline curve: a very rapid log10 reduction on day 1 (first phase) followed by a slow and gradual viral decay starting from day 2 (second phase).3 RBV was shown to have no observable effect on the kinetic parameters such as ϵ or δ of either phase, suggesting that IFN-α dominated or saturated the immune responses in the early phases. If δ = δIFN-α + δRBV, when high-dose IFN-α was used (δIFN-α ≫ δRBV), the immunomodulatory activity will be saturated by that of IFN-α, limiting RBV's contribution to a negligible level. By keeping RBV's immunomodulatory effect to a minimum, the effect in long-term responses may be attributed mainly to the mutagenic effect if any, allowing a more convincing measurement of RBV mutagenesis.

In the Nature paper the authors theorized that the action of RBV was due to lethal mutagenesis within two viral compartments: infectious versus noninfectious virions. A key parameter ρ (related to η), representing the fraction of newly produced noninfectious virions due to RBV action, was termed the effectiveness of RBV mutagenesis.16 Based on this model, total viral production is quantitatively blocked by IFN-α by a factor of (1-ϵ) but is qualitatively reduced by RBV during the next round of infection by a factor of (1-ρ) (Fig. 1). The clinical effect of RBV seems to be more cumulative and delayed with the characteristic “memory” effect that can prevent residual viruses from coming back after cessation of therapy. By factoring in a ρ value, the authors successfully predicted long-term response (ETR and SVR) as a function of IFN-α effectiveness (ϵ) and RBV effectiveness (ρ). When ρ was set at 0.5, the model best predicted both ETR and SVR in a range of IFN-α effectiveness (0.4-0.99) with maximal enhancement by RBV at approximately 25% to 30%, consistent with clinical observations. In contrast to the report by Herrmann et al., the loss rate of infected cells (δ) in this report was not affected by RBV, suggesting that both immune killing (k, mediated by T cells and NK cells) and curing (q, through noncytolytic removal of intracellular viruses) (Fig. 1) were derived mainly from high-dose IFN-α. The effect of RBV on long-term responses predicted by this model derived solely from RBV mutagenesis.16

How does one reconcile this apparently conflicting conclusion by Dixit et al. with that reported by Herrmann et al.? Previously, RBV was seen to have an apparent effect on viral kinetics by some but not by others, which led to a different conclusion regarding the mechanism of action. A closer look at this suggests that the difference may be a function of IFN-α effectiveness. For those studies using high-dose daily IFN-α, the IFN-α effectiveness is generally high (ϵ >0.9); therefore, the immunomodulatory effect of RBV becomes insignificant or undetectable. This was true when Dixit et al. modeled high-dose IFN-α (ϵ = 0.95) with RBV (assuming maximum RBV effectiveness ρmax = 1) or without RBV (ρ = 0): the difference in antiviral effect (log10 reduction) was only about 0.1. For those studies using standard or approved IFN-α doses (IFN-α three times a week or pegylated IFN-α once a week), the IFN-α effectiveness was lower (<0.9, with mean values ranging from 0.36 to 0.67).14 Therefore, the immunomodulatory effect of RBV became apparent, mainly reflected by an increase in the net loss rate of infected cells. Using the model by Dixit et al., RBV could have a significant impact on viral decay, contributing to an additional 2 log10 reduction in viral titer at the end of 2 months (assuming ϵ = 0.5, median δ = 0.14 day−1).16 The observed immunomodulatory activity made it more difficult for Herrmann et al. to draw a definitive conclusion on the mechanism of action between immunomodulation and mutagenesis.14

The state of viral kinetic modeling provides a basic yet incomplete understanding of the underlying processes, but nevertheless allows one to grasp trends and devise upper and lower bounds on virus kinetic parameters.17 Assumptions were made to simplify curve-fitting, but they may not be correct. Furthermore, viral dynamics were built on patients who responded to therapy but not on those who failed to respond. There is more to learn about the dynamic nature of HCV replication in humans. We still do not understand the mechanism of alanine aminotransferase normalization induced by RBV, which hinders the interpretation of the observed immune response. Studies on mechanisms of IFN-α action at the molecular level may also shed light on the interpretation of antiviral kinetics. For example, is ADAR1 induced during IFN-α therapy? Can ADAR1 be mutagenic and introduce mutations into HCV genomes? A ρ value of 0.5 was assumed to predict the ETR and SVR of IFN-α and RBV combination therapy.16 However, at the same RBV dose, the ρ value was much lower to fit the slight viral titer decline observed during RBV monotherapy.10 This gap can only be explained if synergy exists between IFN-α and RBV. Is ADAR1 the missing link for synergy proposed between IFN-α and RBV?

Future modeling should also consider the pharmacodynamic properties of IFN-α and RBV. Early viral kinetics predicted early viral responses well, but its predictive value for long-term response has been somewhat unreliable.3, 18–20 Modeling at later time points (during weeks 6 to 24) is lacking. Late viral kinetics may be more useful to predict treatment success or to optimize therapy, although it may be complicated by patients who achieve early viral responses with undetectable viral titers at later time points. Nevertheless, simultaneous modeling of viral kinetics and drug levels as described by Powers et al.21 may shed light on the temporal effect of dose reduction or dose increase toward treatment outcomes. Clearly there is a need to refine the modeling under real-world treatment settings, where dose reduction may be needed to avoid intolerable side effects or dose increase may be required to improve treatment response. Interestingly, a recent study has suggested that SVR might be further improved when RBV doses were increased up to 4,000 mg daily.22 Can ρ do a better job? The model of Dixit et al. predicts it will.

References

  1. Top of page
  2. Abstract
  3. Comments
  4. References