The emergence of drug resistant viruses, together with the possibility of increased virulence, is an important concern in the development of new antiviral compounds. Cidofovir (CDV) is a phosphonate nucleotide that is approved for use against cytomegalovirus retinitis and for the emergency treatment of smallpox or complications following vaccination. One mode of action for CDV has been demonstrated to be the inhibition of the viral DNA polymerase.

We have isolated several CDV resistant (CDV^R) vaccinia viruses through a one step process, two of which have unique single mutations within the DNA polymerase. An additional resistant virus isolate provides evidence of a second site mutation within the genome involved in CDV resistance. The CDV^R viruses were 3–7 fold more resistant to the drug than the parental viruses. The virulence of the CDV^R viruses was tested in mice inoculated intranasally and all were found to be attenuated.

Resistance to CDV in vaccinia virus can be conferred individually by at least two different mutations within the DNA polymerase gene. Additional genes may be involved. This one step approach for isolating resistant viruses without serial passage and in the presence of low doses of drug minimizes unintended secondary mutations and is applicable to other potential antiviral agents.

We first established the concentration of CDV that would effectively eliminate wt VV plaques without having significant cytotoxic effects on the BSC40 cells (Figure 1). The concentrations that were utilized were based on previous work [9] which indicated that the EC₅₀ for CDV is ~50 μM. Concentrations as low at 50 μM were effective in eliminating plaque formation. We chose to use 150 μM, a value three times the EC₅₀ as the concentration for selection of mutants and no cytotoxicity was apparent at this concentration in these cells. Two different virus strains, VV WR and VV TK::GFP were initially used to isolate resistant mutants. The presence of GFP made the identification of small plaques much easier; however, this parental virus is thymidine kinase (TK) negative thus attenuating the virus and rendering it an unsuitable backbone to later assess the CDV^R virus phenotype in animals [10]. Indeed, since TK indirectly impacts DNA synthesis, a second goal of these studies was to determine whether the inhibitory concentrations of CDV and subsequent mutant selection were impacted by deletion of this enzyme. This was deemed to not be the case as mutants were readily isolated from either virus at comparable concentrations of CDV and is consistent with results published previously [11]. To insure selection of independent mutations, 10 individual plaque purified stocks from both VV WR and VV TK::GFP, were used as the parental lines for the isolation of resistant mutants. A total of six independent resistant viruses were isolated and the DNA polymerase, E9L, gene was sequenced from each virus (Table 1). Each of these viruses contained a mutation(s) in the viral DNA polymerase.

In order to confirm that the mutation detected in the E9L gene was responsible for the CDV resistance, we performed a series of marker rescue experiments and the results of the marker rescue experiments for isolates CDV^R 1 and 2 are shown in Figure 2. DNA fragments from drug resistant isolates were amplified by PCR and used to transfect wild type VV infected cells (Fig. 2A). Resulting viruses were plaqued in the presence of CDV to score for marker rescue (Fig. 2B). Only PCR products that contained a mutation conferring resistance to CDV should produce plaques, in this example, PCR fragments E9 and 14 (Fig. 2B). To remove the possibility of second site mutations in the original virus, resistant viruses were reconstructed in a wild type VV background by transfecting PCR products containing only a single mutation in E9L. For CDV^R 1 and 2 we used PCR fragment 14 containing the A314V mutation. For CDV^R15 a fragment containing the M671I mutation was used and for CDV^R16, a fragment with the ΔK174 mutation was transfected. The reconstructed viruses are designated with an "A" following the original virus name to distinguish them from the original isolates. Reconstructed viruses containing only the identified mutation in E9L in a wild type VV background were sufficient to confer CDV resistance except for CDV^R 15. We were unable to reconstruct the CDV^R 15A virus (Table 1) cleanly despite several attempts and the resulting reconstructed virus always contained two mutations within the E9L gene, the original mutation (M671I) and a second mutation that corresponds to the same mutation found in CDV^R 16 (ΔK174). This was observed multiple times and indicates quite clearly, that the resistance that allowed the isolation of CDV^R15 containing the M671I mutation depends on a second contributing mutation in the original CDV^R 15 isolate outside of the DNA polymerase for resistance. Furthermore, this second site mutation in CDV^R15, outside the DNA polymerase can be compensated for by a specific second mutation in the DNA polymerase, ΔK174. In contrast, CDV^R 16A was successfully reconstructed to contain only the original ΔK174 mutation.

Each of the three reconstructed viruses, CDV^R 1A, 15A and 16A were analyzed for their growth properties compared to wt VV. The growth curves in Figure 3 indicate that all three CDV resistant strains grew less well than wild type virus although CDV^R 16A did produce titers reaching wild type levels after an initial lag in growth. All three resistant strains produced very small plaques compared to wild type virus, with CDV^R 1A producing "pinpoint" plaques after three days. CDV^R 1A and 15A ultimately produced much less total virus than wild type or CDV^R16A.

We confirmed that the viruses were resistant to CDV in two additional cell lines at another laboratory (Table 2). The EC₅₀ for CDV was obtained in HFF and Vero cells and compared to two independently obtained strains of parental VV WR. The resistant viruses had EC₅₀ values that were 3 to 7 fold higher than the parental virus strains. The greatest resistance was with CDV^R1A containing the mutation at A314V which produced the smallest plaques and lowest titers.

It has been previously reported that CDV^R VV is attenuated in mice. We assessed the virulence of our reconstructed viruses in mice inoculated intranasally (Tables 3 and 4) and confirmed that all of our resistant strains were significantly attenuated in this model. No mice were killed by the CDV^R15A strain even at the highest dose given, so an LD₉₀ could not be established.

No crystal structures of VV E9 exist in the protein database. However, E9 exhibits significant homology to the type B family of DNA polymerases. Residues 429–809 of E9 could be aligned with residues 279–597 of Thermostable B Type DNA Polymerase from Thermococcus gorgonariu with 41% homology (E value = 1e-15). This region represents the catalytic core of the DNA polymerase. By modeling E9 on the crystal structure of Thermococcus gorgonariu DNA polymerase (PDB code 1TGO) it appears that the location of the M671I mutation is not far from the active site in the putative polymerase domain (Figure 4). A summary of known mutations conferring drug resistance is presented in Figure 5. A number of mutations cluster in the exonuclease domain, including those at residue 314 as previously noted [5].

The results obtained from these studies provides further evidence that the primary but not sole target of CDV is the viral DNA polymerase and that drug resistance can be a significant problem even in the presence of relatively low doses of drug. It is important to note that these drug resistant mutants all had reduced virulence in mice and suggest that the development of these mutants may not contribute to enhanced disease.

Monolayer cultures of BSC40 cells (Dr. Richard Condit) were maintained in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum (FBS) (Gibco), 50 IU of penicillin, and 50 μg of streptomycin per ml (Cellgro, Herndon, Va.) [12]. CV1 cells (ATCC, CCL-70) were maintained in minimal essential media (MEM) with Earle's salts supplemented with 5% FBS, 340 mM sodium pyruvate, 50 U/ml penicillin, 50 μg/ml streptomycin and non-essential amino acids. Vero Cells were obtained from ATCC and were maintained in MEM with Earl's salts and the addition of 10% FBS and standard concentrations of L-glutamine, penicillin and gentamicin. Methods for obtaining and passaging human foreskin fibroblast (HFF) cells were described previously [13]. All cell lines were maintained at 37°C in the presence of 5% CO₂.

The parental viruses used for isolation of CDV^R mutants were VV WR and VV TK::GFP. VV TK::GFP contains the GFP gene driven by the synthetic VV early-late promoter inserted into the thymidine kinase (tk) gene. This virus was generated via standard methods using a pSC65GFP clone in order to recombine the GFP gene into the TK locus of wild type VV [14]. Virus titers were determined by standard plaque assays. To obtain CDV^R mutants, 20 independent virus stocks were generated from single plaques from the original VV TK::GFP (lines 1–10) and VV WR (lines 11–20) virus stocks.

CDV was provided by Gilead Sciences, Foster City, CA. Stock solutions of CDV (5 mM) in DMEM without serum was stored at 4°C and protected from light.

Confluent monolayers of BSC40 cells in 6-well plates were infected with 2 × 10⁴ PFU/well of either VV WR or VV TK::GFP in DMEM with no supplements except for 150 μM CDV. After 60 min of adsorption an additional 1.5 ml of DMEM with 10% FBS, antibiotics and 150 μM CDV was added to each well. Plates were incubated at 37°C for 48 h and examined for plaques under the light microscope, or with fluorescence for GFP containing plaques. To isolate identified plaques, the liquid medium was carefully removed and the plaque was scraped with a 1 ml large bore pipette tip and transferred into 1 ml of DMEM without serum and stored at -80°C. Routinely, 2 plaques from each individual virus stock were isolated. The virus from the original plaque was plaque purified one additional time under agarose and in the presence of 150 μM CDV to ensure that it was a single isolate. From these dishes, plaques were picked and subsequently amplified.

DNA sequences of the DNA polymerase (E9L) gene from the viruses were obtained by direct sequencing of the PCR products amplified from the total DNA of infected cells. The DNA was prepared from virus infected cells with the DNeasy Tissue Kit (Qiagen Inc., Valencia,, CA), according to the manufacturer's protocol. The entire E9L gene was PCR amplified using two primers IDT327, 5'-ATGGATGTTCGGTGCATTAATTGGT-3' and IDT328, 5'-TTATGCTTCGTAAAATGTAGGTTTTGAACC-3' and then sequenced with primers that hybridize within E9L to give overlapping sequence data. Sequencing was performed by the University of Florida ICBR DNA Sequencing Core Laboratory.

Mapping of the individual mutations conferring resistance and the reconstruction of the mutation(s) in a wild type VV background were performed as described previously [15]. Confluent monolayers of BSC40 cells in 6 well dishes were infected with 5 × 10³ PFU of wt VV WR in a volume of 0.5 ml and 30 min later transfected with 1.5 – 2 μg DNA complexed with 12 μl Lipofectamine 2000 per manufacturer's instructions (Invitrogen). Different PCR products from the mutant CDV^R viruses were used for transfection including products containing the entire E9L gene or products containing only portions of E9L gene. A map of the fragments used is found in Figure 2. PCR fragment 13 is approximately 5 kb and contains approximately half of E9L at the 3' end as well as DNA downstream of E9L into E6 (primers 5'-TACGATGTTGTAAAGTGTACGAAGCG-3'; 5'-AGTTAGAGAAATGACGTTCATCGGTG-3'). The 5' portion of E9L is contained in a 5 kb fragment, #14, generated with 5'-TTTGTTTTGGAGCAAATACCTTACCG-3' and 5'-CGAGAGTGGTTGAATGTTTGACTGTG-3'. As a negative control, fragment 15 approximately 2.9 kb upstream of E9L translational start site was used in transfections (5'-AAATAGTCACGCAATTCATTTTCGGG-3'; 5'-TGCTTTTGATGGTAATTTCTGGTGCC-3'). All primers and fragment numbering is from Luttge and Moyer, 2005 The cells were incubated at 37°C for 1 h while rocking, then an additional 2 h without rocking. DMEM containing 150 μM CDV was added to each well. Plates were incubated at 37°C for 48 hr, and then harvested. The resulting viruses were grown on BSC40 cells in the presence of 150 μM CDV. When mapping mutations, the dishes were stained with crystal violet. For reconstruction CDV^R plaques were picked from the plaques obtained from the infection/transfection mixture and later amplified on BSC40 cells for stocks. The E9L gene of the reconstructed viruses was sequenced as described above and compared to the original mutation from which it was derived.

BSC40 cell monolayers in 60 mm dishes were infected with 200 PFU of VV WR, VV TK::GFP or CDV^R mutant virus suspension in 0.5 ml DMEM without serum and with CDV concentrations of 0, 50, 150, 250 or 500 μM. After adsorption for 60 min at 37°C, the medium was carefully aspirated and the wells were overlaid with 1% agarose mixed with an equal volume of 2 × DMEM, 10% FBS and the same concentration of CDV used during the initial infection. The plates were incubated for 4 days at 37°C and then stained with 0.26% crystal violet in 10% ethanol, 22% formaldehyde and plaques were counted.

HFF cells were added to 6-well plates two days prior to the assay. On the day of assay, drug at two times the final desired concentration was diluted serially 1:5 in 2× MEM with 10% FBS to provide six concentrations. Culture medium was aspirated from triplicate wells for each drug concentration and 0.2 ml per well of diluted virus was added which yielded 20–30 plaques per well. The plates were incubated for one h with shaking every 15 minutes. Equal volumes of 1% agarose and drug solutions were mixed and added to each well in 2 ml volumes and the plates incubated for three days. Cell monolayers were stained with neutral red and plaques were enumerated using a stereomicroscope at 10× magnification. 50% effective concentration (EC50) values were calculated by standard methods.

Growth properties of the reconstructed CDV^R viruses were compared to growth of wild type VV. BSC40 cells (1 × 10⁵) in twelve well dishes were infected individually with each virus, wt VV, CDV^R 1A, CDV^R 15A, CDV^R 16A, at an MOI = 0.01 PFU/cell, in duplicate. After 1 h adsorption, the virus was removed and the cells washed with PBS. One ml of DMEM containing 10% FBS was added to cells. Infected cells were incubated at 37°C and harvested by scraping at the following time points: 1, 3, 6, 9, 12, 18, 24, 48, 72 h post infection. The virus was released from the cells by 3 freeze thaw cycles and titered on CV1 cells.

Female BALB/c mice, 3 weeks of age, were obtained from Charles River Laboratories, Raleigh, North Carolina. Mice were group housed in microisolator cages and utilized at a quantity of 10–15 mice per group. Mice were obtained, housed, utilized and euthanized according to USDA and AAALAC regulatory policies. All animal procedures were approved by University of Alabama at Birmingham, Institutional Animal Care and Use Committee prior to initiation of studies. BALB/c mice were anesthetized with ketamine-xylazine prior to virus inoculation. VV infections were initiated by intranasal inoculation of media containing varying concentrations of wild type and drug resistant mutants of VV ranging from 8 × 10⁵ to approximately 1 PFU/animal, depending on the titer of each virus stock. Virus suspension was instilled into both nostrils using a micropipetor and a total volume of 40 μl per animal. For these experiments mice were checked for mortality at least once daily for 21 days, but twice daily during the period when peak mortality was expected to occur. The mortality observed for the wild type virus, such as VV WR, was compared with that observed with the CDV^R VV.

The amino acid sequence of E9 was aligned with that of Thermostable B Type DNA Polymerase from Thermococcus gorgonariu (derived from PDB file 1TGO) by Blast. A homologous model was calculated based on the amino acid sequence alignment and the known structure 1TGO using Modeller (version 9.2) [16]. The model structure was displayed by PyMol (Delano Scientific, San Carlos, CA).