In Vitro-Selected Drug-Resistant Varicella-Zoster Virus Mutants in the Thymidine Kinase and DNA Polymerase Genes Yield Novel Phenotype-Genotype Associations and Highlight Differences between Antiherpesvirus Drugs

Varicella-zoster virus (VZV), one of the eight human herpesviruses, is classified together with herpes simplex virus types 1 (HSV-1) and 2 (HSV-2) in the subfamily Alphaherpesvirinae. Following primary infection, VZV causes varicella (chickenpox), characterized by a generalized vesicular rash. Varicella, one of the most contagious infectious diseases in humans, is mainly a childhood disease in unvaccinated populations. Although varicella is usually a mild to moderate illness in immunocompetent patients, primary or secondary immunodeficiency is associated with serious complications, such as central nervous system involvement, pneumonia, and secondary bacterial infection, that can lead to the death of the infected patient (22). Older age and possibly pregnancy are also important risk factors associated with severe varicella and fatal outcome (59).

Subsequent to primary infection, the virus remains latent in the dorsal root ganglia of the host, and reactivation may occur several years or decades later, resulting in herpes zoster (HZ), also known as shingles. A painful vesicular rash with a dermatomal distribution characterizes HZ. Postherpetic neuralgia (PHN), the persistence of pain after resolution of the HZ rash, is the most troublesome complication associated with HZ. Worldwide, HZ affects millions of individuals, and a decline in VZV-specific cell-mediated immunity (CMI) because of aging or immunosuppression increases the risks of VZV reactivation (54, 55, 64).

A live-attenuated vaccine is available for the prevention of varicella, and its use in the pediatric population since 1995 has resulted in a significant reduction in the incidence of varicella. However, in rare circumstances in immunosuppressed children who were undiagnosed at the time of vaccination, the vaccine has caused severe infections or disseminated disease that required prompt and prolonged treatment with antivirals (10, 17, 30, 35). A zoster vaccine is also approved by the Food and Drug Administration (FDA) for prevention of HZ in adults 60 years of age and older (19). The vaccine has proven to be safe and partially effective in preventing both HZ and PHN (20, 34, 46).

Varicella is generally a self-limiting disease, and in most cases, symptomatic treatment is sufficient (59), but specific antiviral therapy for varicella is required for groups of people at risk for severe disease (22). Therapy for HZ is warranted to accelerate healing, diminish the duration of acute and chronic pain, and reduce the risk of PHN. Additionally, in immunocompromised patients, antiviral therapy for HZ is needed to limit the risk of dissemination of VZV (1, 68). Acyclovir (ACV), valacyclovir (an oral prodrug of ACV), and famciclovir (an oral prodrug of penciclovir [PCV]) are approved for the treatment of VZV infections. Brivudin (BVDU; bromovinyldeoxyuridine) has been licensed for the therapy of HZ in some European countries. Both brivudin and its arabinofuranosyl analogue sorivudine (BVaraU) are contraindicated in patients with cancer because they can be metabolized to the free base (E)-5-(2-bromovinyl)uracil, which blocks the degradation of the anticancer drug 5-fluorouracil (5-FU), resulting in the accumulation of high (toxic) 5-FU levels (9).

Although ACV resistance (ACV^r) is virtually unobserved in immunocompetent patients, it can be a major problem among immunosuppressed hosts, mainly those who have received prolonged ACV therapy (25, 36, 42, 58). Approximately 30% of VZV infections clinically resistant to ACV are associated with virological resistance (our unpublished data). A lack of clinical response to ACV has been correlated with the presence of thymidine kinase (TK)-deficient or TK-altered virus, although a few cases were documented with an alteration in DNA polymerase (Pol) (18). Thus, the molecular basis for resistance to ACV is linked to mutations in either the viral TK or the DNA Pol gene. Herpesvirus TKs are able to transfer a γ-phosphoryl group from ATP to the 5′ hydroxyl group of thymidine (13). In addition, VZV and HSV-1 TKs possess thymidylate kinase activity, which converts dTMP to dTDP. Then, the conversion to dTTP is carried out by cellular kinases. Herpesvirus TKs have a broader substrate specificity than mammalian cytosolic TKs and can phosphorylate many nucleoside analogues, such as ACV, PCV, GCV, BVDU, BVaraU, and the bicyclic pyrimidine nucleoside analogues (BCNAs), a new class of extremely potent and specific anti-VZV agents (5, 38). The difference in substrate specificity between cellular and viral TKs is the basis for the selectivity of nucleoside analogues as antiherpetic compounds.

Because most ACV^r VZV strains have alterations in the viral TK (31, 33, 41, 47, 49, 53), foscarnet (PFA) (a direct inhibitor of viral DNA Pol) is the drug of choice to treat ACV^r VZV infections (1, 11, 21, 32, 48). A few reports have described VZV mutants with alterations at the level of DNA Pol associated with resistance to PFA (26, 66, 67). The acyclic nucleoside analogues (ANPs), such as cidofovir, which do not require activation by the virus-encoded TK, can be considered alternatives for the treatment of ACV^r and/or PFA^r mutants, as shown in the cases of HSV-1 and HSV-2.

In several studies, laboratory and clinical ACV^r VZV mutants were reported (8, 14, 28, 62). However, investigations correlating a detailed phenotype with specific mutations in the VZV TK and DNA Pol genes are lacking. The analysis of drug resistance with VZV is more complex than with HSV because VZV is normally cell associated, with only a very small fraction of cell-free virus, making it difficult to isolate pure populations of viruses. In the present study, we were interested in genotyping previously isolated drug-resistant VZV emerging under selective pressure with different classes of antiherpesvirus agents, including ACV, PCV, BVDU, BVaraU, four different BCNAs, the pyrophosphate analogue PFA, and two ANPs (i.e., 2-phophonylmethoxyethyl [PME] purine derivatives) (2, 3). Furthermore, the contributions of specific mutations in VZV TK or DNA Pol to antiviral drug resistance and their impacts on the structures of viral proteins were investigated. Our results revealed patterns of cross-resistance associated with specific mutations in VZV TK and DNA Pol and highlight important differences, not only between distinct classes of antivirals, but also between members of one defined structural class of drugs (i.e., ACV and PCV).

In the first part of this study, we correlated specific mutations in VZV TK with detailed phenotyping and analyzed the impacts of amino acid substitutions associated with drug resistance on the structure of the viral enzyme. We showed here that the genetic basis for resistance to the BCNAs, as well as to ACV, BVDU, and BVaraU, was alterations in the viral TK gene. Mutants arising under selective pressure with four different BCNAs were investigated, including Cf1743, whose 5′ valyl ester (i.e., FV-100) has been evaluated recently in a clinical phase 2 trial (4, 38, 43). The data show a therapeutic benefit of FV-100 given at 200 or 400 mg/day in reducing the severity and duration of shingles-related pain, the incidence of postherpetic neuralgia, and the time to lesion healing. A phase 2b trial is envisioned to further explore its antiviral efficacy. In cell culture, BCNAs display highly potent anti-VZV activity and have an absolute requirement for VZV TK-mediated activation for antiviral activity. They were shown to be efficiently phosphorylated by VZV TK, but not by the TKs of other herpesviruses, such as HSV-1 and HSV-2 (6). Obligate intracellular 5′ monophosphorylation of Cf1743 is required for biological activity (37).

Among 67 plaque-purified clones analyzed for alterations at the level of VZV TK (emerging under ACV, BVDU, BVaraU, and the four different BCNAs), 55 showed an insertion or a deletion of a C in a string of 6 Cs at nucleotides 493 to 498 of the viral TK gene. An insertion or a deletion in this run of Cs was found in 8 out of 10 independent selection procedures. Hence, this homopolymer region appears to function as a hot spot within the VZV TK gene, producing nonfunctional, truncated TK proteins. This run of 6 Cs at nucleotides 493 to 498 in the VZV TK gene is not found at homologous positions in the HSV-1 or HSV-2 TK gene. Two other frameshift mutations in VZV TK were due to an insertion of 2 nucleotides at position 412 or to a T deletion at nucleotide 641. These changes did not occur in homopolymeric repeats of VZV TK. It is intriguing to observe that the major predominant resistance mechanism against VZV TK for the VZV TK-specific BCNAs is the 493 to 498 C insertion that is also observed for BVaraU and ACV. Given the high selectivity of BCNAs for VZV TK, but not HSV TK, one might have expected unique amino acid mutations that do not occur with other antiherpesvirus drugs in VZV TK or in the HSV-1 TK.

In contrast to VZV, sequencing of the TK genes of HSV from in vitro-isolated and clinical ACV-resistant mutants revealed the presence of several homopolymeric nucleotide (G or C) stretches that function as hot spots (16, 27, 40, 51, 52, 56). Approximately 50% of the cases of ACV resistance in HSV are due to nucleotide insertions or deletions in homopolymers of Gs or Cs in the viral TK gene, mainly in the two longest homopolymers, one of 7 Gs and one of 6 Cs. It appears that the TK genes of high-GC-content herpesviruses, such as HSV (GC content, 68%) and bovine herpesvirus type 1 (BHV-1) (GC content, 72%), contain several stretches of Gs or Cs where insertions or deletions of nucleotides lead to frameshift mutations (39, 51). VZV, which has a markedly lower GC content (46%) than HSV or BHV, has only a few homopolymer stretches (three of 5 As, one of 5 Ts, and one of 6 Cs). According to our results, the stretch of 6 Cs at nucleotides 493 to 498 emerged as the most important hot spot in the VZV TK gene for insertions or deletions. Hot spots for mutations in canine herpesvirus, another herpesvirus with a low GC content (32%), has also been shown to occur in only two stretches of 8 As in the viral TK gene (69).

In a recent study, the proportions of short sequence repeats (SSRs), which include minisatellites, microsatellites, and homopolymers, were compared among different herpesviruses and mimivirus (61). Although larger SSRs are more obvious to the eye, homopolymers are the most abundant class of SSRs in all viral genomes analyzed thus far. SSRs are likely to be a common property of large DNA viruses, and the authors suggest that homopolymers across the viral genome are mutational hot spots for evolutionary diversity in herpesviruses. VZV and HSV homopolymer regions are likely sites of mutation, both within the TK gene and in the other genes. This study also points to a higher average protein-coding variation for HSV-1 than for VZV: 1.3% versus 0.2%, respectively. The higher variation in HSV-1 correlated with a higher G/C percentage overall (68% for HSV-1 versus 46% for VZV) and in SSRs (84% for HSV-1 and 47% for VZV). Furthermore, a comparison of interstrain variation in homologous proteins of HSV-1, VZV, and pseudorabies virus (PRV) across 3 HSV-1, 3 PRV, and 18 VZV strains highlighted several proteins that appear to vary more substantially in one virus than another. Interestingly, UL23 (HSV-1 thymidine kinase) showed a higher percentage of amino acid variation (1.33%) than the homologs in VZV (ORF36) and PRV (UL23) with, respectively, 0.39% and 0.31% amino acid variation across strains.

Another remarkable difference between the HSV and VZV TKs is the smaller number of genetic polymorphisms in VZV TK than in HSV-1 and HSV-2 TKs (44). It appears that VZV TK is relatively conserved, and only a very low number of genetic polymorphisms have been described (i.e., L288S and S179N). In our study, the T86A amino acid change in VZV TK always occurred in combination with a deletion or an insertion of a C in the homopolymeric stretch of 6 Cs (nt 493 to 498). Considering the high degree of conservation of VZV TK, the location of Thr-86 in a region where mutations have been associated with drug resistance, and its effect on the structure of the viral protein, it can be assumed that the T86A change may affect the activity of the enzyme. Two other novel mutations were found in VZV TK that could be linked to drug resistance: E192-stop (selected under pressure with BVDU) and the G24R substitution in the ATP-binding site (upon selection with one of the BCNAs). The importance of the changes at position G24 in VZV TK is highlighted by its impact on the structure of the viral enzyme. Furthermore, in a previous work, Boivin et al. characterized the G24E mutation in the ATP-binding site of VZV TK, and the phenotype of the clinical strain bearing this mutation was shown to be TK deficient (8). Mutations at homologous residues in HSV-1 and HSV-2 (i.e., G61) have also been linked to drug resistance (44).

Although ACV resistance in HSV and VZV may involve different mechanisms (i.e., reduction or loss of viral TK activity, alteration of substrate specificity for viral TK, and mutations affecting DNA Pol), most of the ACV^r HSV and VZV strains are TK deficient (18). Our studies are in agreement with these findings (Table 1). Such frameshift mutations, as also observed to occur in the VZV TK gene under BCNA drug pressure, resulted in poor, if any, phosphorylating activity of the mutant TK enzyme, as examined in the extracts of HEL cell cultures infected with mutant VZV strains that were selected in the presence of Cf1603 or Cf1743 (data not shown). In contrast, the PCV^r viruses had a mutation at the level of the DNA Pol gene. Furthermore, Ida and collaborators found that the emergence frequency of resistant VZV mutants was significantly higher following ACV exposure than following PCV treatment (24). No data are available regarding the isolation of PCV-resistant mutants in the clinic. This can be explained by the longer time of exposure to PCV required for selection of resistance compared to ACV and by the type of mutants isolated, since TK is not vital for the virus while only a limited number of mutations that are nonlethal for the virus can occur in viral DNA Pol.

The differences between HSV and VZV regarding selection of ACV^r and PCV^r mutants can be explained either by the different natures of the viruses or by subtle structural differences between PCV and ACV. First, there is a higher number of homopolymer stretches in HSV TK than in VZV TK, and no long homopolymers of Gs are found in VZV TK. VZV TK contains only one run of 6 Cs that is a hot spot for insertion or deletion of G. Second, the triphosphate form of ACV (ACV-TP) acts as an obligate DNA chain terminator when incorporated into the growing DNA strain, while the triphosphate form of PCV (PCV-TP) can be incorporated into the growing DNA chain, resulting in limited chain elongation (65). Third, the affinities and therefore the molecular interactions of ACV, PCV, and their triphosphates with the viral TK and DNA Pol differ. In herpesvirus-infected cells, the efficiency of phosphorylation to the triphosphate derivatives is higher for PCV than for ACV. Instead, ACV-TP is a more potent inhibitor of HSV-1 and VZV DNA polymerases than PCV-TP (12). Fourth, the triphosphate forms of ACV and PCV differ in their intracellular half-lives in infected cells (much longer for PCV-triphosphate than for ACV-triphosphate). Fifth, it has been proposed that ACV-TP may be capable of modifying the enzymatic or proofreading properties of HSV DNA Pol to selectively slip or stutter in regions containing G-C homopolymers and induce frameshift mutations (50, 51). Although the K_i of PCV-TP with DNA Pol is significantly higher than that of ACV-TP, PCV-TP is actually a more efficient and complete DNA chain terminator under physiological conditions (with competing natural nucleotides present) than is ACV-TP (12, 45). In conclusion, the more complete DNA chain termination activity of PCV-TP, the weaker interaction with DNA Pol than ACV-TP, and the lack of homopolymeric stretches of Gs in the VZV TK and Pol genes may explain the differences observed between ACV and PCV in the present study. It is worth mentioning that different mutational patterns in TK genes of HSV were reported with viruses selected in cell culture for resistance to ACV or PCV (50, 60). Additionally, PCV, similar to GCV, may be phosphorylated, not only by VZV TK, but also by the protein kinase ORF47, which is homologous to the HCMV protein kinase UL97.

VZV encodes a protein kinase (ORF47) that is homologous to the HCMV protein kinase UL97. As HCMV does not encode a TK, the UL97 protein kinase is responsible for GCV activation in infected cells. Koyano et al. (29) analyzed the phosphorylation pathways of antiherpesvirus nucleoside analogues by VZV-specific enzymes. Cells expressing VZV TK proved to be susceptible to ACV and GCV, as well as oxetanocin analogues. In contrast, VZV PK-expressing cells were susceptible only to GCV and oxetanocin analogues, indicating that VZV PK phosphorylates certain nucleoside analogues, for example, GCV and oxetanocin. Further evidence for the activation of GCV and oxetanocin analogues was provided by the fact that these compounds were able to inhibit the replication of TK-negative VZV strains at concentrations 10 times higher than those at which they inhibited wild-type VZV. These results are in agreement with our phenotyping data for mutant VZVs bearing different types of changes in the viral TK gene, which showed an increase in EC₅₀s of only 0.7 to 1.0 log unit compared to 1.5 to 1.8 log units for ACV and PCV and >2 log units for BVDU and the BCNAs (Fig. 4).

In the present study, we also investigated the DNA Pol changes in drug-resistant virus mutants. Since the PCV^r plaque-purified isolates did not show alterations at the level of the TK gene and showed cross-resistance to the pyrophosphate analogues and the PME derivatives, we performed genetic analysis of the DNA Pol gene. PCV seemed to select for virus strains that contain mutations in VZV DNA Pol (i.e., V666L) that were also reported to occur under PFA pressure (66, 67). Among the four novel mutations (i.e., D668Y, L767S, M808V, and M874I) in the viral DNA Pol that could be linked to drug resistance, two positions (i.e., Met-808 and Leu-767) are conserved among HSV, VZV, and HCMV, while two are partially conserved (i.e., Asp-668 and Met-874). Their effects on the structure of VZV DNA Pol (Fig. 6 and 7) highlight the impact of mutations at these positions.

A recent study has highlighted the limitations of employing the crystal structures of related polymerases, such as that of the bacteriophage RB69 or Thermococcus gorgonariusas, as model systems to explain mechanisms of inhibition of DNA synthesis and drug resistance (63). It was demonstrated that the RB69 DNA polymerase (gp43) is ∼400-fold less sensitive to the pyrophosphate analogue PFA than the HCMV DNA polymerase (UL54). These data point to profound differences in sensitivity and selectivity to antiviral drugs when the viral enzyme is compared with gp43. Critical regions of the nucleotide-binding site of gp43 were replaced by equivalent regions of the HCMV enzyme. Interestingly, the chimeric gp43-UL54 enzymes containing residues of the helix N and helix P of UL54 were resensitized to PFA and ACV. Hence, chimeric enzymes can provide a valuable tool to study the mechanisms of action and resistance to PFA and to nucleotide analogue inhibitors.

It should be noted that mutations in viral TK are usually associated with high levels of resistance (≥1 log₁₀-fold increase in EC₅₀s). In contrast, mutations in viral DNA Pol result in smaller differences in EC₅₀s between the wild-type and mutant viruses.

Our results illustrate the complexity of the mechanisms of herpesvirus resistance to antiviral drugs. Although there is conservation among viral TKs and DNA Pols, marked differences between VZV and HSV could be found. In addition, differences between HSV and VZV regarding the types of mutants selected under ACV and PCV pressure were observed. The relevance of these findings to the in vivo situation warrants further investigation.

This work was supported by the Belgium Federal Public Service “Public Health, Food Chain Safety and Environment, action 29 of the National Cancer Plan”; by the Fonds voor Wetenschappelijk Onderzoek (FWO) Krediet no. G-0608-08; and by the Program Financing (PF-10/18) of the K. U. Leuven.

We are grateful to Anita Camps, Lies Van Den Heurck, Steven Carmans, and Lizette van Berckelaer for excellent technical assistance.