The exact mechanism of the retroviral restriction remains elusive. Although some degree of retroviral restriction can be obtained in transfection assays using active-site mutants of APOBEC3G, suggesting the existence of a deamination-independent restriction pathway (4, 8, 17, 22, 23, 29, 38, 46, 51), APOBEC3G mutants which lack deoxycytidine deaminase activity give a substantial reduction in the efficiency of restriction (5, 13, 22, 32, 34, 35, 37, 40, 45). This major deaminase-dependent pathway of restriction is associated with G-to-A hypermutation of the retroviral genome (19, 32, 33, 53). However, the restriction is unlikely to simply be the consequence of the accumulation of a high mutation load since, although dependent on the integrity of APOBEC3G's catalytic site (Fig. (Fig.1A),1A), efficient restriction of human immunodeficiency virus (HIV) (ΔVif) particles is evident even with low mutation loads, such as those produced in presence of APOBEC3C (Fig. (Fig.1B1B).

It was initially suggested that restriction might result from the uracilation of the retroviral cDNA (16, 21, 32, 33). However, it is notable that although some reports have indicated a role for the UNG as well as possibly the SMUG1 uracil DNA glycosylase (42, 50), APOBEC3-mediated restriction is actually unaffected by a deficiency of UNG activity (24, 34). Consistent with this result, we find that expression of Ugi, a potent and irreversible inhibitor of UNG, in human 293T cells does not influence the antiviral properties of APOBEC3G on HIV ΔVif infection whether the inhibitor is expressed in the producer cells and/or in the target cells (Fig. (Fig.1C).1C). Sequence analysis of integrated HIV DNA also revealed no significant difference in the accumulation of G→A mutations in viruses produced in presence or absence of Ugi (Table 1). However, although UNG is an evolutionarily ancient uracil-DNA glycosylase and is implicated in excising uracils that arise in DNA through dUTP misincorporation (as well as during antibody gene diversification), SMUG1 appears to be the major uracil DNA glycosylase responsible for excising uracils that arise in cellular DNA from cytidine deamination (2). Indeed, SMUG1 could be a strong contender, since it is capable of excising uracils from both single-stranded and double-stranded uracilated DNA (2, 39). Thus, the lack of effect of UNG deficiency on APOBEC3G-mediated HIV restriction does not exclude the possibility that uracil excision is important to the process. Indeed, Fig. Fig.1D1D depicts a biochemical assay in which a double-stranded oligonucleotide (5′-ATTATTATTATTCCUGGATTTATTTATTTATTTATTTATTT-fluorescein-3′ and 5′-AAATAAATAAATAAATAAATAAATCCAGGAATAATAATAAT-3′) containing a central U·A pair and a fluorescein marker at its 3′ end was incubated with whole-cell extracts of Ugi-expressing 293T transfectants. The results reveal the presence of substantial residual uracil excision activity; incubation with the SMUG1-inhibiting antibody Pms-1 (25) shows this residual activity to be due to SMUG1.

We were therefore interested in ascertaining whether APOBEC3G-mediated restriction would be affected by simultaneous deficiencies in both UNG and SMUG1. Although SMUG1-deficient mammalian lines have not been described, we have previously found that chicken cells do not exhibit SMUG1 activity or indeed any major UNG backup activity as judged by biochemical assays of cell extracts (9-11). We therefore asked whether human APOBEC3G could restrict an avian retrovirus in chicken cells.

Stocks of replication-competent Rous sarcoma virus (RSV) in which the src gene had been replaced by an enhanced green fluorescent protein (eGFP) reporter (41) were used to infect either chicken fibroblast DF1 cells or DF1 transfectants that stably expressed human APOBEC3G. Supernatants from these infected cell populations were then used to infect either DF1 control cells or DF1[APOBEC3G] transfectants; eGFP expression was monitored by flow cytometry 24 h later (Fig. (Fig.2A).2A). Expression of human APOBEC3G in the viral producer cells yielded viruses that exhibited an approximately 80% reduction in viral infection as judged by eGFP fluorescence. As with APOBEC3G-mediated restriction of HIV type 1 infectivity in human cells (44), the restriction depends upon APOBEC3G expression in the producer cells rather than in the targets. Similarly, viral encapsidation of human APOBEC3G was detected in purified virions produced from APOBEC3G-expressing DF1 cells (Fig. (Fig.2B).2B). It is intriguing that a mammal-specific retroviral restriction factor is packaged even by an avian retrovirus, suggesting that it might exploit some ancestral pathway to force itself into the virus; binding to conserved cellular cofactors, such as the 7SL RNA, could provide such a mechanism (26, 47, 48). The restriction is also associated with retroviral deamination. Target DF1 cells were collected and integrated viral sequences analyzed for mutations. The vast majority of mutations were G-to-A transitions as read on the plus-strand viral DNA (Table 1), with a local sequence specificity characteristic of human APOBEC3G (preference for Cs at positions −1 and −2 relative to the target cytosine) (Table 2) (18- 20, 33).

We then investigated whether inhibition of uracil excision activity in DF1 cells affected APOBEC3G restriction of RSV. Introduction of a Ugi-expressing vector into DF1[APOBEC3G] cells yielded stable transfectants that exhibited no detectable uracil excision activity as judged by biochemical analysis of cell extracts using an oligonucleotide cleavage assay (Fig. (Fig.2C).2C). Viral infectivity carried out using these cells showed that the Ugi-mediated neutralization of UNG activity in the viral producer cells had no effect on APOBEC3G-mediated viral restriction (Fig. (Fig.2A).2A). We also did not observe any significant difference in viral restriction if Ugi was expressed in the target (as opposed to producer) cells (data not shown). The Ugi-mediated inhibition of uracil excision in DF1 also had no effect on the extent or nature of the retroviral mutations that accompanied the restriction (Tables 1 and 2). Furthermore, not only is APOBEC3G-mediated restriction in DF1 cells not diminished by inhibiting UNG activity, the restriction also is not enhanced by ectopic expression of SMUG1 (Fig. (Fig.2A2A).

These results reveal that restriction by APOBEC3G requires neither UNG nor SMUG1. Therefore, the dependence on the integrity of the catalytic site for the major pathway of APOBEC3-mediated restriction reflects either (i) a need for APOBEC3 to recognize cytosine but not necessarily for it to deaminate cytosine, (ii) that the deamination-induced alteration of particular nucleotide sequences within the retroviral genome compromises specific interactions necessary for retroviral replication and/or genomic integration (an explanation we think unlikely in view both of the low mutation load that accompanies restriction as well as the fact that different restricting APOBEC3 family members exhibit different local sequence preferences), or (iii) that restriction does occur through recognition of the APOBEC3-generated uracil but this recognition is not attributable to UNG or SMUG1 and might or might not involve base excision. Although thymine-DNA glycosylase is another glycosylase capable of excising uracil from DNA (14), it seems to us an unlikely candidate since it excises uracil only from U·G mispairs, which would arise only from cytidine deamination occurring following retroviral second-strand DNA synthesis. However, apart from the possibility of an involvement of an as-yet-unidentified mammalian uracil excision enzyme, it is also conceivable that restriction could be mediated by uracil recognition without excision. For example, it has recently been shown that archaebacteria harbor an endonuclease (which displays regions of homology to mammalian apurinic/apyrimidinic endonuclease 1) that cleaves DNA adjacent to uracil (15). A similar activity has been identified in flies (3). It will obviously be interesting to ascertain whether such activities can be found in mammalian cells.

Finally, it is notable that while high-level expression of APOBEC3G has been shown to be capable of inhibiting TY element transposition in yeast (13, 43), the restriction of retroviral infection achieved by expressing human APOBEC3G in chicken DF1 cells is strikingly efficient. APOBEC3 proteins appear to be restricted to mammals, and so, consistent with a recent report by Wiegand and Cullen (49), it appears that the evolution of APOBEC3 proteins was sufficient to create an effective pathway of retroviral restriction with all the factors necessary for incorporating APOBEC3 into the retrovirus as well as for the mediation of the downstream pathway of restriction already being in existence prior to mammalian evolution.