The threat that smallpox could be used against military and civilian populations has prompted national investment in the development of new antipoxviral therapies. The major pharmaceutical companies are not inclined to initiate searches for new drugs to combat a disease that exits only in theory and for which the best possible outcome would be that the drug is never used. However, the success of SIGA in discovering ST-246, a new antipoxviral agent suitable for smallpox prophylaxis (41
), signals that there is an opportunity for niche companies to do good while doing well. Academic and public-sector scientists can also play a role in antipoxviral discovery, especially in the identification and dissemination of information about new inhibitors that could be of use to the research community as probes of poxvirus biology. The route to discovery of novel agents by screening off-patent medicines approved for human use has been advocated as a means to find “new tricks for old drugs” (39
) that can bypass the laborious and costly preclinical steps and go straight to clinical trials. This approach is of little interest to the private sector, which is focused on new entities that yield proprietary compositions of matter.
Our screen of 2,880 compounds was “validated” by the output of 13 compounds that inhibited poxvirus replication. The confirmed hits included an agent with known antipoxvirus activity (mycophenolic acid), one related to a known inhibitor (ancitabine), and inhibitors of eukaryal protein synthesis that have a favorable therapeutic index (cycloheximide, anisomycin, emetine, and cephaeline). The antipoxviral activities of cardiac glycosides, lycorine, and mitoxantrone are new findings.
Mitoxantrone is unique among the inhibitors identified here in that it has no major impact on viral gene expression. Viral late protein synthesis proceeds normally in the presence of mitoxantrone, but there is clear biochemical and microscopic evidence of a mitoxantrone-induced block to virion assembly. The antiviral target is likely to be a viral protein, because we were able to isolate mitoxantrone-resistant viruses by passage in the presence of drug. To our knowledge, this is the first instance in which a useful bioactivity of mitoxantrone in eukaryotic cells depends on a target protein other than topoisomerase II.
Our results highlight a “new trick” for mitoxantrone, but it is not the first time that a clinically useful drug with a known mechanism has exerted a highly specific effect on poxvirus replication by a different mechanism. Rifampin, an antibacterial drug used for the treatment of tuberculosis, is an inhibitor of bacterial DNA-dependent RNA polymerase, and yet rifampin's activity against poxviruses entails a block to virion morphogenesis with no effect on viral gene expression (52
Mitoxantrone now joins several other selective antipoxvirus agents that affect virus assembly: novobiocin, rifampin, TTP-6171, ST-246, and IMCBH. Novobiocin inhibits a very early stage of morphogenesis, so that membrane crescents and spherical immature virions do not accumulate (46
). Rifampin causes assembly to arrest after the formation of abnormal membrane-enclosed structures (rifampin bodies), which contain viroplasm but lack the rigid spherical shape of normal immature virus particles (52
). Rifampin resistance and hypersensitivity are attributable to point mutations in the D13 gene (2
). TTP-6171 targets the I7 protease and causes a morphogenetic arrest similar to that seen in an I7 temperature-sensitive virus at a restrictive temperature (8
). IMCBH and ST-246 inhibit the release of infectious progeny virus particles from the cell by preventing the wrapping of IMV by Golgi membranes (40
). IMCBH and ST-246 target the F13 gene product, a 37-kDa protein component of the Golgi membrane-derived viral envelope (44
The results of whole-genome sequencing pinpoint vaccinia DNA ligase as the target of mitoxantrone's antipoxviral activity. Three different MX-R strains have three distinct mutations that result in either amino acid substitutions (C11Y or C145Y) or a single extra amino acid insert within the N-terminal domain of the 552-amino-acid viral ligase polypeptide. This portion of vaccinia DNA ligase is located well upstream of the active site of nucleotidyl transfer and is dispensable for the ligase adenylylation step of the enzyme's nick-sealing reaction (47
). Deletion of the N-terminal domain of vaccinia ligase (from amino acids 1 to 200) slows the rate of strand joining and lowers the enzyme's affinity for nicked duplex DNA (47
). The Cys11 and Cys145 residues, which are determinants of sensitivity to mitoxantrone, are both strictly conserved in the DNA ligases encoded by nearly all other vertebrate poxviruses, including variola, monkeypox, ectromelia, myxoma, deerpox, and swinepox. The two notable exceptions are the fowlpox and canarypox DNA ligases, in which the equivalent of Cys11 is conserved but the amino acid corresponding to vaccinia Cys145 is replaced by histidine and tyrosine, respectively. Thus, it is possible that canarypox might be naturally impervious to mitoxantrone by virtue of its mimicry of the C145Y ligase mutation that renders vaccinia mitoxantrone resistant.
Although it is not obvious how these DNA ligase mutations could protect vaccinia against mitoxantrone, we are struck by the consilience of our findings with those of DeLange et al. (12
), who reported that (i) vaccinia virus replication in BSC40 cells was inhibited by etoposide, another clinically effective anticancer drug that acts as a topoisomerase II poison, (ii) etoposide had no apparent effect on viral gene expression, and (iii) an etoposide-resistant vaccinia virus isolated from a mutagenized stock after six rounds of passage in the presence of etoposide had a resistance-conferring C11Y mutation in the N-terminal domain of vaccinia DNA ligase. We can extrapolate from the published data an IC99
value of ~100 μg/ml etoposide (diluted from a solution of pure drug in DMSO) for suppression of progeny virus yield in a synchronous infection (12
). This dose corresponds to a concentration of ~130 μM etoposide (molecular weight of 589 g/mol), which explains why our high-throughput screening assays for >90% inhibition of vaccinia replication at 10 μM concentrations of test compounds did not identify etoposide as a hit, even though it was present in both the Prestwick and Microsource compound libraries. Mitoxantrone is apparently more potent than etoposide in blocking vaccinia virus replication.
The C11Y vaccinia ligase mutation that sufficed to confer resistance to etoposide and cross-resistance to amsacrine (another topoisomerase II poison) (12
) is identical to the MX-R3 mutation that confers resistance to mitoxantrone. DeLange et al. speculated that the replication block imposed by etoposide might be caused by a defect in forming the hairpin telomeres at the ends of the viral genome (12
). However, the observed decrement in the level of monomeric viral genomes in the presence of etoposide seemed insufficient to account for the reduction in the yield of infectious progeny. DeLange et al. did not assess the effects of etoposide on virus morphogenesis or processing of virion structural proteins. In the case of mitoxantrone, the discrete block to virus assembly readily accounts for the reduction in virus yield.
Although mitoxantrone, etoposide, and amsacrine are all topoisomerase II poisons, they have significantly different chemical structures. Neither etoposide nor amsacrine is an anthracenedione (like mitoxantrone). It is not clear whether the antipoxviral activity of these compounds stems from (i) their similar effects on cellular topoisomerase II, which could result in a pretoxic lesion that somehow depends on vaccinia DNA ligase for its antiviral manifestation, or (ii) a common property of these compounds not related directly to topoisomerase II, e.g., their capacity to intercalate into or otherwise bind DNA. At least in the case of etoposide and amsacrine, the antiviral effects cannot be attributed simply to inhibition of vaccinia DNA ligase activity, insofar as deletion of the nonessential vaccinia ligase gene confers resistance to these two drugs, albeit not to the same degree as the C11Y mutation (12
). Pulse-field electrophoretic analysis of vaccinia DNA in etoposide-treated infected cells showed the accumulation of subgenomic DNA fragments that were not evident in the absence of the drug (12
), consistent with cellular topoisomerase II having access to replicating viral DNA and breaking it in the presence of the drug poison. It is notable that these viral DNA fragments are still observed in etoposide-treated cells infected with the etoposide-resistant C11Y ligase mutant (12
), which implies that the drug-induced breakage of viral DNA does not by itself block production of infectious progeny virus. Further studies of the replication of wild-type and mitoxantrone-resistant viruses in the presence and absence of drugs are needed to fully elucidate the basis for the mitoxantrone-induced morphogenetic block.
Finally, this study demonstrates the power of whole-genome sequencing to rapidly locate poxvirus mutations that cause phenotypes of interest. The classical approach to map drug resistance phenotypes entails the use of genomic fragments from the resistant poxvirus strain in a transfection-based assay that scores for marker uptake by the sensitive parental strain to yield drug-resistant progeny (2
). This method requires fragment library construction from each virus mutant and is considerably more laborious and time consuming than sequencing and analyzing the genome of a mutant virus, which can be accomplished in several days. The potential drawback of the whole-genome approach is that the analysis could reveal multiple nucleotide changes in different genes that still need to be sorted out by marker rescue before the resistance locus can be assigned. In the present study, we found that the genome sequence of our wild-type vaccinia WR was surprisingly stable compared to that of the reference strain. Moreover, the mitoxantrone-resistant strains accrued amazingly few incidental changes after multiple passages and plaque purification. Thus, we could confidently assign the mitoxantrone resistance phenotype to DNA ligase mutations based on the fact that every mitoxantrone-resistant virus had a mutation in the ligase gene and that two of the resistant strains had no nucleotide changes elsewhere in their genomes.