Didanosine, interferon-alfa and ribavirin: a highly synergistic combination with potential activity against HIV-1 and hepatitis C virus

Objective: To evaluate the antiviral triple combination didanosine (ddI), interferon-alfa (IFN-α), and ribavirin for potential synergy in inhibition of HIV-1 replication in vitro.

Methods: Phytohaemagglutinin-stimulated cord blood mononuclear cells were infected with HIV-1_IIIB or the HXB2D molecular clone of HIV-1 then cultured with interleukin-2 with ddI, ribavirin or IFN-α, alone and in combination. Reverse transcriptase activity was measured after 7 days to determine the inhibitory concentration of 50% (IC₅₀) for the various drugs in replicate assays. Analysis of combined effects was performed using both the median effect principle (CalcuSyn, Biosoft^®) and three-dimensional modelling (MacSynergy II).

Results: The triple combination was highly synergistic against HIV-1 in vitro with combination indices < 1. The mean IC₅₀ was reduced from 6.85 to 0.90 μmol/l (P < 0.001) for ddI and from 6.58 to 1.00 μmol/l (P < 0.001) for IFN-α. No increased cytotoxicity was observed. Results were similar with both viral strains and using both analyses. In the triple combination, increasing concentrations of IFN-α resulted only a slight enhancement of synergy: synergy volumes were 134 [95% confidence limit (CL), 77-191] with 5 U IFN-α and 214.92 (95% CL, 116-314) with 10 U. This supporting the observation that the majority of the synergistic activity was derived from the combination of ddI and ribavirin, with IFN-α providing additional additive suppression.

Conclusions: This novel triple combination has the potential to provide simultaneous activity against both HIV and hepatitis C and deserves further study to determine if can be safely administered in the clinical setting.

Introduction

As the nature of the HIV epidemic changes, an increasing number of persons infected with HIV are also chronically infected with hepatitis C virus (HCV). Despite the significant impact of triple combination antiretroviral therapy (HAART) on the control of HIV, no beneficial effect on HCV replication has been observed in co-infected patients [^1,2]. In fact, therapy for HIV may adversely affect HCV-related liver disease, perhaps because of restored anti-HCV responses and/or direct hepatotoxicity [³]. Increasingly, there is a recognized need to treat HCV, often in the context of background antiretroviral therapy. Information is, therefore, needed on the interactions that may occur between drugs used to treat both infections.

The standard treatment for HCV is combination therapy with interferon-alfa (IFN-α) and ribavirin. Interestingly, both IFN-α and ribavirin have anti-HIV activity in vitro. IFN-α demonstrates anti-HIV activity probably through several mechanisms. In vitro, it interferes with viral assembly and budding in T cells and protects macrophages from infection [⁴]. IFN-α is also synergistic in vitro with reverse transcriptase inhibitors such as zidovudine (AZT) and didanosine (2′,3′-dideoxyinosine; ddI), presumably by acting at different viral targets [⁵]. Various formulations of IFN-α are currently in clinical trials for HIV treatment. Ribavirin, a purine analogue, is a broad-spectrum antiviral agent. Ribavirin has been shown to suppress HIV viral replication in T cells in vitro at concentrations of 50 μg/ml or higher by interfering with post-transcriptional processing [^6,7]. Of interest, ribavirin is synergistic at much lower concentrations with other purine analogues such as ddI [^8,9]. Ribavirin increases the intracellular phosphorylation of ddI by inhibiting inosine monophosphate dehydrogenase, resulting in increased concentrations of inosine monophosphate, which can then act as a phosphate donor for the reaction ddI to ddI monophosphate [¹⁰]. Conversely, ribavirin antagonizes the activity of AZT through inhibition of AZT phosphorylation [¹¹].

Treatment of HIV comprises various triple combinations of antiretroviral drugs generally from at least two different classes. Successful combinations have been shown to be highly synergistic in vitro, which may explain their potency [¹²]. Therefore, existing drugs used in the treatment of HIV and HCV, respectively, may act synergistically when combined, raising the prospect of treating both infections simultaneously. In these studies, we tested this hypothesis by evaluating the antiviral triple combination of ddI, IFN-α and ribavirin for potential synergy in inhibition of HIV-1 replication in vitro.

Materials and methods

In vitro drug susceptibility testing was performed by a modified AIDS Clinical Trials Group Department of Defence consensus method as described below.

Drugs and viral isolates

Ribavirin and ddI were purchased from Sigma (Mississauga, Ontario, Canada). Human interferon-alfa A (IFN-α) was obtained from PBL Biochemical Labs, Inc. (New Brunswick, New Jersey, USA). AZT was purchased from Sigma for use as a negative control. Viral isolates used were HIV-1 strain IIIB (HIV-1_IIIB) or the HXB2D molecular clone of HIV-1.

Cell culture and viral infection

Human cord blood mononuclear cells (CBMC) were obtained from the Department of Obstetrics of the Jewish General Hospital and extracted using Ficoll-Hyaque density gradient centrifugation (Amersham Pharmacia Biotech AB, Uppsala, Sweden). CBMC were cultured in a base culture medium consisting of RPMI-1640 (Gibco BRL, Life Technologies, Paisley, Scotland), 10% heat inactivated fetal bovine serum (FBS; Gibco BRL), penicillin (100 U/ml), streptomycin (100 μg/ml) and 2 mmol/l glutamine, as described previously [¹³]. CBMC were stimulated for 3 days with 0.1% phytohaemagglutinin (PHA) and maintained at 37°C and 5% CO₂. PHA-stimulated CBMC were infected with virus at a multiplicity of 2 × 10⁸ median tissue culture infective dose per ml (TCID₅₀/ml) for 2 days or 2 × 10⁹ TCID₅₀/ml for HIV-1_IIIB for 2 h (10⁶ cells/ml). Cells were then washed once and resuspended in base medium further supplemented with 5% (v/v) interleukin-2 (Boehringer Mannheim GmbH, Germany) and hydrocortisone (Abbott Laboratories, St. Laurent, Quebec, Canada) and immediately plated as described below to conduct experiments.

Drug stock solutions

Stock solutions for ddI and ribavirin were prepared by dilution in sterile distilled water and stored at -20°C. AZT was prepared as a stock in dimethyl sulfoxide. The stock concentrations were 2.5 mg/ml for ddI, 10 mg/ml for ribavirin and 10 mg/ml for AZT. IFN-α was supplied at a concentration of 5 × 10⁷ U/ml in a solution of sterile phosphate-buffered saline and 0.1% bovine serum albumen, from which dilutions were made directly. On the day of each experiment, serial dilutions for each drug were freshly prepared in base medium.

Susceptibility and cytotoxicity assays

At least six serial dilutions of ddI, ribavirin or IFN-α and a drug-free control were used. Freshly infected CBMC were placed in 96-well flat bottom plates at 4 × 10⁵ cells/well in 124 μl medium, and then drugs were distributed to achieve a final volume of 248 μl/well. After 3 days incubation, half of the supernatant was discarded and each well was replenished with appropriate concentrations of drugs and interleukin-2. After 7 days incubation, cytotoxic effects (cytotoxic concentration for 50%; CC₅₀) of drugs were assessed by measuring inhibition of proliferation of exponentially growing cells. The 50% inhibitory concentrations (IC₅₀) of each drug alone was performed by measuring the reverse transcriptase activity in culture supernatant through the incorporation of [³H]-dTTP (NEN, Boston Massachusetts, USA) by using poly(A).oligo(dT) (Pharmacia, Mississauga, Ontario, Canada) as template-primer, as previously described [¹⁴]. Viral reverse transcriptase activity was counted using the Wallac-1410 liquid scintillation counter.

Once the IC₅₀ for individual drugs was determined, double combinations were set up in a checkerboard fashion. For example, ddI was plated in serial dilutions vertically and ribavirin or IFN-α in serial dilutions horizontally. Triple combinations were studied by using a double checkerboard with and without a fixed concentration of a third drug added to all wells (overlay drug). By altering the amount of the overlay drug, various concentrations were tested. The percentage inhibition of viral replication was calculated for each combination by measuring the reverse transcriptase activity compared with their respective controls. Experiments were repeated at least three times and with both viral isolates to validate results.

Data analysis and evaluation of synergy

Analysis of synergy, additive or antagonist effects of the antiviral agents was first performed according to the median effect principle using the CalcuSyn computer program [^14,15] to provide estimates of the IC₅₀ values of drugs alone and in combination. Mean IC₅₀ values obtained with single drugs were compared with the respective IC₅₀ values obtained in combinations using two-sided Student's t-tests with α = 0.05 level of significance. This method of Chou and Talalay has been shown to be equivalent to Lowe additivity [¹⁶]. Combination indices (CI) are estimated from the data and reflect the nature of the interaction between drugs. CI values < 1 represent synergistic activity; CI = 1 are additive and CI > 1 represents antagonism. The value of CI is directly proportional to the amount of synergy in the combination regimen. For example, values < 0.5 represent a high degree of synergy whereas values > 1.5 represent significant antagonism. This approach has been widely used in the analysis of antiviral interactions and was chosen to allow comparability with published literature. This method requires combining drugs in fixed ratios. The following ratios were used: ddI to ribavirin, 1 : 10; IFN-α to ddI, 1 : 0.1; IFN-α to ribavirin, 1 : 1.

In other experiments, data were analysed using the Pritchard and Shipman's MacSynergy II software (kindly provide by M. Pritchard) using the independent effect method [¹⁷]. This program allows the three-dimensional examination of drug interactions with Bliss Independence as the null reference model using a Microsoft Excel-based interface. All data points generated from the checkerboard are used in the analysis. Confidence bounds are determined from replicate data. If the 95% confidence limits (CL) do not overlap the theoretic additive surface, then the interaction between drugs differs significantly from additive. When positive (above the additive surface) the interaction is synergistic; when below the surface, the interaction is antagonistic. The volumes synergy or antagonism can be determined, graphically depicted in three-dimension, and represent the relative quantity of synergism or antagonism produced per change in the two drug concentrations [¹⁷].

Results

Activity and cytotoxicity of individual agents

The mean (±SD) IC₅₀ values obtained in three independent experiments were ddI, 6.85 ± 2.02 μmol/l for ddI, > 50 μmol/l for ribavirin, and 6.58 × 2.19 U for IFN-α. No cytotoxicity was observed with ddI or IFN-α at the maximal doses tested (12.5 μmol/l ddI and > 250 U/ml IFN-α). Ribavirin caused increasing inhibition of cellular proliferation at doses > 50 μmol/l, with average CC₅₀ of 100 μmol/l.

Synergistic action of didanosine and ribavirin

Ribavirin markedly potentiated the activity of ddI in vitro. The combination resulted in a 10-20-fold decrease in the average IC₅₀ of ddI, from 6.85 to 0.38 μmol/l (P = 0.08) with HIV-1_IIIB. Furthermore, the IC₅₀ of ribavirin in the combination was reduced to 4.43. The combination was highly synergistic with CI values of ≪ 1 (Table 1). Similar results were obtained using the HXB2D strain (mean IC₅₀ of ddI in the combination 0.36 μmol/l; CI values 0.038-0.432). Three-dimensional analyses confirmed the synergistic interaction between the two drugs (Fig. 1). Synergistic interactions were observed across the entire concentration grid with a calculated synergy volume of 678.71 (95% CL, 504-853; Fig. 1b). No increase in cytotoxicity was observed over the entire spectrum of drug concentrations tested. As a negative control, AZT and ribavirin were combined and were found to be antagonistic at most concentrations tested, with a mean CI of 2.42 (Table 1).

Fig. 1

Table 1

Synergy between didanosine and interferon-α

Synergism was also observed in the combination of ddI and IFN-α. The mean IC₅₀ with HIV-1_IIIB was reduced from 6.85 to 0.174 μmol/l (P = 0.003) for ddI and from 6.58 to 1.90 μmol/l (P = 0.04) for IFN-α. Similar reductions in IC₅₀ was observed with HXB2D strain (mean IC₅₀ for ddI in the combination was 0.31 μmol/l). For both viral strains, the CI values were well below 1: 0.169-0.477 for HIV-1_IIIB and 0.05-0.61 for HXB2D. Three-dimensional analyses confirmed the synergistic interaction and the absence of antagonism (Fig. 2). The synergy volume was notably lower with this combination, at 202.78 (95% CL, 23-383), compared with that observed with ddI and ribavirin. The majority of the interaction was best characterized as additive, with synergy occurring at lower drug concentrations (0.001-0.5 μmol/l ddI and 1-5 U IFN-α; Fig. 2b).

Fig. 2

Activity of ribavirin and IFN-α against HIV

The combination of IFN-α and ribavirin was more complex that that seen with other combinations tested. The interaction between these drugs was best characterized as additive (CI values just below 1) but was mildly synergistic against HIV replication at lower concentrations (see Table 1) and demonstrated some antagonism at higher drug concentrations (Table 1). In the MacSynergy II analysis, antagonism was seen at lower drug concentrations (synergy volume -289.4; 95% CL, -59 to -520) with a small degree of synergy at mid-concentrations (synergy volume 4.36; 95% CL, 0-9) and the remainder of the interaction best characterized as additive (Fig. 3).

Fig. 3

The triple combination enhances activity against HIV

Triple combinations were first examined by studying a double checkerboard of ddI and IFN-α with ribavirin (10 μmol/l) as the overlay drug. The concentration of ribavirin tested was chosen to reflect its highest level of activity in double combination with ddI below that where cytoxicity was observed. A further reduction of the IC₅₀ values of both ddI and IFN-α was observed in the presence of ribavirin. The average IC₅₀ values of ddI and IFN-α were further reduced from 6.85 to 0.09 and from 6.58 to 1.0, respectively (P < 0.001). The majority of synergy appeared to be derived from the interaction of ddI with ribavirin, as there was no statistically significant change in the synergy volumes in the triple combination compared with those seen with the double combination of ddI and ribavirin (data not shown). To assess the contribution of IFN-α to the activity of the triple combination further, double checkerboards of ddI and ribavirin with IFN-α as the overlay drug were studied. Various concentrations of IFN-α were tested (1-10 U). This design permitted the evaluation of a full range of ddI, ribavirin and IFN-α concentrations. Increasing concentrations of IFN-α resulted only a slight enhancement of synergy [i.e. synergy volume 134 (95% CL, 77-191) with 5 U IFN-α and 214.92 (95% CL, 116-314) with 10 U; Fig. 4], suggesting again that the majority of the synergistic activity in the triple combination was derived from the combination of ddI and ribavirin. Again no increase in cytotoxicity was observed and results were similar using the two HIV strains.

Fig. 4

Discussion

The novel triple combination of ddI, IFN-α and ribavirin was found to be highly synergistic against HIV-1 in vitro. The IC₅₀ of ddI in the triple combination was reduced 80-fold compared with ddI alone. Given the known activity of IFN-α and ribavirin against HCV, these findings raise the prospect of a new clinical strategy for the treatment of dual infection.

These studies extend understanding of the previously reported synergy between ddI and ribavirin. CI values were far below 1 (0.008-0.47), indicating a high degree of synergy was present. This synergy translated into up to a 20-fold reduction in the IC₅₀ of ddI. Furthermore, we have characterized for the first time the nature of the interaction surface using detailed three-dimensional analyses. Significant synergy was present across the entire concentration grid. Thus, even at low concentrations of ribavirin, well below cytotoxic levels, considerable enhancement of ddI activity is observed. This powerful interaction between ddI and ribavirin has been attributed to the inhibition by ribavirin of inosine monophosphate dehydrogenase, which results in increased intracellular levels of ddI monophosphate [¹⁰]. This interaction could lead to enhanced anti-HIV potency in vivo. In the only published clinical trial of ddI and ribavirin therapy, ribavirin was added after 4 weeks of ddI monotherapy. A significant decline in HIV viral load was observed after 12 weeks. However, because of the study design, it was not possible to compare the effect of the combination relative to ddI monotherapy [¹⁸].

The in vitro combination of IFN-α and ddI was also synergistic, with CI values ranging from 0.17 to 0.48. Previous reports have shown synergy of IFN-α with AZT and ddI against clinical isolates obtained from individuals with prior AZT monotherapy, with CI values in similar range [¹⁹]. In the three-dimensional analyses presented here, the synergy volumes observed were modest compared with those seen between ddI and ribavirin and occurred predominantly at the lower concentration ranges of the two drugs. Therefore, the majority of the interaction between ddI and IFN-α was additive and likely results from the two drugs acting at independent targets rather than an enhancement of the intracellular activity of ddI, which occurs with ribavirin.

We report for the first time on the combination of IFN-α and ribavirin against HIV. The interaction between these drugs was predominately additive, with mild synergy at lower drug concentrations (CI values 0.59-1.3). Use of MacSynergy analyses highlighted the complexity of the drug interaction, which demonstrated additive effects and antagonism depending on the mixture of drug concentrations. These findings emphasize the difficulties in interpreting drug interactions in vitro, particularly when testing agents that have multiple mechanisms of action.

Finally, the triple drug combination resulted in even more pronounced inhibition of HIV-1 in vitro, giving an 80-fold reduction in the IC₅₀ of ddI. An examination of the synergy volumes suggested that the bulk of the synergy observed came from the interaction between ddI and ribavirin, with IFN-α providing additional additive suppression.

Considerable debate exists in the literature about the best methods for analysing additivity in drug combinations [^16,17]. The two major competing models are termed the Lowe additivity and the Bliss independence null reference models. Furthermore, while two drug analyses are commonly performed, the evaluation of triple combinations is even more complex and difficult to depict graphically. We chose to use two methods to analyse our data, each using one of the above definitions of additivity. The median effect principle of Chou and Talalay was selected primarily because it had been used widely in the literature and, therefore, provides a means of comparison with previously published data on interactions between antiretroviral drugs. It has been shown to be equivalent to Lowe additivity [¹⁶]. This method has some important limitations, which have been reviewed particularly for mutually non-exclusive inhibitors [^16,17]. Importantly, it requires the use of fixed drug concentration ratios, which limits the examination of drug interactions across a diagonal line of the dose-response interface and may miss complexities in drug-drug interactions that might be observed at other concentration mixtures. For these reasons, a three-dimensional model has been developed and is easily applicable to analyse replicate data using simple computer software. The MacSynergy II program of Prichard and Shipman follows a Bliss independence model, allows for the analysis of the entire interaction surface and provides 95% CL values.

In these multiple replicate experiments, the analyses using the two models yielded very similar results, strongly supporting that the synergy between drugs reported did indeed exist. Furthermore, the three-dimensional analysis provided greater insight into the nature of the drug interactions across all concentrations tested and demonstrated that synergy can be achieved at clinically achievable drug concentrations.

Although in vitro synergy does not necessarily translate into in vivo potency, there are several examples where this is the case. Notably, the triple antiretroviral combination of indinavir, AZT and lamivudine has resulted in the long-term suppression of HIV in infected individuals [²⁰]. This triple combination has been shown to be highly synergistic in vitro, using analyses similar to those reported here [¹²]. Like the interaction seen in the present study, the majority of synergy occurred between lamivudine and AZT. The interaction between indinavir and AZT was only mildly synergistic and was best characterized as additive. The addition of clinically achievable concentrations of lamivudine markedly improved synergy.

An important consideration with respect to the potential utility of ddI and ribavirin combinations in clinical practice is safety. One relevant example is the combination of ddI and hydroxyurea. Like ribavirin, hydroxyurea is a potent inhibitor of cellular ribonucleotide reductases. In vitro studies have demonstrated that the combination of ddI and hydroxyurea is synergistic both against wild-type strains of HIV and against several clinical isolates with ddI resistance [²¹]. Clinical studies have supported this [²²]. However, concerns about increased toxicity with the combination of hydroxyurea and ddI have hampered enthusiasm for this approach in treating patients [²³]. It is important to recognize that the combination of ddI and ribavirin may likewise enhance ddI-related toxicity. There are few clinical data on this combination. The only published trial of ddI and ribavirin in therapy for HIV showed no increase in pancreatitis or peripheral neuropathy [¹⁸]. However, recent reports of lactic acidosis in subjects receiving ddI, stavudine and ribavirin [^24,25] highlight that careful evaluation of new antiviral combinations is required prior to adopting such therapies into clinical practice.

The lack of an in vitro culture system for HCV precludes antiviral susceptibility studies for this agent. Clinically, however, the combination of IFN-α and ribavirin is far superior to treatment with either agent alone, suggesting the presence of additive or synergistic activity. In fact, ribavirin has no clinical activity against HCV, and the action of IFN-α is modest at best. When combined, the long-term response rates in chronic HCV increase to 38-49%, compared with 5-13% with IFN-α alone [^26,27]. Furthermore, the combination can achieve long-term responses in patients who have previously failed IFN-α monotherapy, perhaps demonstrating activity against resistant viral strains [²⁸].

Currently there is no standard approach to the management of HIV and HCV co-infection. Problems faced by the co-infected population are numerous. Co-infected patients experience higher rates of hepatotoxicity because of their chronic hepatitis [^29,30]. Eradication of HCV may, therefore, promote tolerance of HAART. Response rates to HCV treatment in the setting of co-infection are reduced with lower CD4 cell counts [³¹] suggesting that HIV treatment should not be withheld. Injection drug use and chaotic lifestyles makes adherence to the complex regimens currently used to treat HIV and HCV difficult. The ideal regimen in theses circumstances, therefore, should provide simultaneous control of both HIV and HCV, be highly active, minimize the number of pills to be taken and reduce hepatotoxicity. The novel triple combination of ddI, ribavirin and IFN-α has the potential to fulfill this role and deserves further study to determine if can be safely administered in the clinical setting.

Acknowledgments

We thank Maureen Oliviera for valuable technical assistance, Bonnie Spira for administrative support and Melanie Lemay for manuscript preparation

Sponsorship: This work was supported by the Canadian Foundation for AIDS Research (Glaxo-Wellcome BioChem Pharma Positive Action Grant 0120410), the Canadian HIV Trials Network and the Canadian Institutes of Health Research (awards to MBK and MAW).

Note: This work was presented in part at the Eighth Annual Conference on Retroviruses and Opportunistic Infections, Chicago, February 2001 [abstract 308].