Antiviral therapy of chronic HBV infections presents a novel problem. At the beginning of treatment, every hepatocyte is apparently infected by the virus (4
). Moreover, the rate of turnover of this cell population, even with active liver disease, is low (34
> 1 week and, in “healthy” carriers, probably greater than 1 month). Thus, merely inhibiting virus replication would not readily eliminate the virus unless cccDNA, the template for viral RNA synthesis, had a short half-life within infected cells. However, this issue is still controversial. Some studies suggest that the DNA may have a high turnover rate (12
). On the other hand, data from the present and other studies suggest that this DNA is highly stable in vivo (14
). In particular, these data suggest that if cccDNA has a finite life time, its half-life in the chronically infected liver is similar to that of infected hepatocytes.
One mechanism that would accelerate virus clearance is loss of cccDNA during cell division. In the present study, we sought indirect evidence for loss of cccDNA during mitosis by assaying for declines in the average cccDNA copy number in infected cells during therapy with the nucleoside analog L-FMAU. If this DNA is lost during mitosis and if it does not have any intrinsic instability in nondividing cells, then once virus DNA replication is blocked, the cccDNA copy number in infected cells should, ideally, remain fixed as the liver proliferates. That is, cells would either have lost cccDNA through the process of cell division or retained the original amount because they had not yet divided. The data suggest, however, that the cccDNA was distributed to daughter cells during proliferation of infected hepatocytes, producing the observed decline in the average copy number among cells that remained infected after prolonged therapy. Moreover, loss by mitosis requires that the infected-cell number decline virtually from the beginning of therapy, a possibility inconsistent with the experimental findings. The decline in copy number, by itself, could be explained by a model in which cccDNA is lost during mitosis and is also lost by decay in cells that have survived without division throughout the course of therapy. For instance, the observed results at 30 weeks of therapy could be modeled by an infected-cell death rate of 0.75% per day and a cccDNA half-life of 70 days. However, this model predicts that only 60% of the cells would remain infected after 6 weeks of therapy, a possibility at odds with the overall data.
An alternative possibility is that the low copy number was the result of new infections of cells that had lost existing cccDNA in the presence of L-FMAU. If so, the prevalent cccDNA in the liver might then have a drug resistance genotype. However, after 30 weeks of therapy, at which time the average cccDNA copy number among infected cells had declined at least 5- to 10-fold, the wild-type virus sequence was still prevalent in the cccDNA population. Our data thus favor but do not prove the hypothesis that cccDNA survives through mitosis and is distributed to each daughter cell, resulting in a decline in cccDNA copy number per cell. Data from a recent study (13
) of WHV cccDNA survival in primary hepatocyte cultures that were induced to undergo limited proliferation by addition of epidermal growth factor were also consistent with this possibility.
Examination of the data in Fig. and in a previous study (45
) revealed an unexpected result. The type I mutation was sometimes detectable as a prevalent species in serum virus at early times in therapy, when virus titers were still declining. Since the same mutation may be associated with the later rebound of virus titers (Fig. ) (45
) and since the type I mutation confers L-FMAU resistance on a laboratory strain of WHV (Fig. ), the reason for the continued decline at early times is not obvious. Several possibilities, not necessarily mutually exclusive, need to be considered. First, additional mutations outside the sequenced region of the polymerase may contribute to mutant fitness. This was not evident in a previous study, in which the complete pol
gene of selected type I mutants was sequenced (45
). However, the possibility has not been ruled out. Second, the type I mutant may be a common quasispecies that is generated as a result of errors during reverse transcription of a pregenomic RNA that was transcribed from wild-type cccDNA. In that case, it would be expected that virus titers would continue to decline until a significant fraction of this mutant virus could be converted to cccDNA. Third, the type I mutant may have a low replication rate, which, together with the need for coinfection with a virus producing the viral envelope proteins, may delay its spread to uninfected hepatocytes.