Our results demonstrate that APOBEC3G is able to block DND1 function and restore miRNA mediated inhibition of translational repression. This function of APOBEC3G appears to apply to multiple mRNA targets of DND1 as APOBEC3G has a similar effect on P27, LATS2 and CX43. On the other hand, DND1 does not appear to affect the viral restriction function of APOBEC3. In addition, our genetic crosses suggest that DND1 interaction with APOBEC3 does not regulate the mutator function of APOBEC3 in germ cells.
The mechanism of how APOBEC3G blocks DND1 remains to be determined. At least, three possible mechanisms can be proposed. The first possibility is that APOBEC3G may bind to DND1 and sequester it away from mRNAs and miRNAs (Figure 3f). In support of this, we have observed that mouse APOBEC3 and DND1 co-immunoprecipitate and co-localize in cells .
The second possibility is that APOBEC3G may bind mRNAs (maybe together with DND1) and subsequently interact with components of the miRISC to activate translation repression or interact with translation initiation factors to inhibit them. Indeed, mass spectrometric analysis show that a large number of cellular RNA-binding proteins associate with APOBEC3G [40–42]. Some of these are known components of the miRISC such as ARGONAUTE 1(Ago1), ARGONAUTE 2 (Ago2), GW182, MOV10, YB-1, DCP1A and RCK/P54, and are involved in post-transcriptional silencing of gene expression [38, 41–43]. These interactions of APOBEC3G with RNA binding proteins were found to be either direct protein-protein interactions or mediated by RNA. In addition, confocal microscopy experiments showed that APOBEC3G co-localized with many of the miRISC RNA-binding proteins to mRNA processing, P-bodies [40, 41]. Thus it is conceivable that interaction of APOBEC3 with specific miRISC proteins may override the effect of DND1 to enhance miRNA activity.
The third possibility is that the cytidine deaminase activity of APOBEC3 may allow it to edit the 3′-UTR sequences of P27, LATS2 and CX43 to inhibit DND1 binding. Interestingly, indirect support of this hypothesis comes from recent reports by a number of groups that analysed deep sequencing data and found greater than expected incidence of editing present in the mammalian transcriptome [44, 45]. As to which of the three possible mechanisms apply to APOBEC3G blocking DND1 function is currently under investigation.
APOBEC3 proteins have been studied as factors that restrict viruses and retrotransposons. However, we found that DND1 has no effect on the viral restriction function of APOBEC3. APOBEC3G can bind both cellular RNAs and RNA binding proteins [27, 40, 41] and the RNA binding activity of the N-terminal cytidine deaminase domain of APOBEC3G is essential for viral restriction [23, 24]. The significance of APOBEC3G interactions with cellular proteins and RNAs is not clear. However, during viral infection of cells, such as HIV-1 (Vif) infection of T lymphocytes, APOBEC3G gains access to viral particles through a ribonucleoprotein interaction and thus APOBEC3 binding to RNA is a critical for antiviral function . The incorporation of APOBEC3G into new viral particles allows it to be released into infected cells where APOBEC3G can deaminate the replicating viral cDNAs to effect reduction of viral infectivity. Interestingly, it has been shown that other P-body proteins, such as MOV10 [46–48] are also involved in reducing the infectivity of exogenous retroviruses and retrotransposons. Thus, P-body protein components such as MOV10, and as we report here, APOBEC3G, participate both in miRNA silencing and viral restriction processes.
An earlier study on APOBEC3G function showed that APOBEC3G can, by itself, inhibit miRNA activity . The previous study used luciferase constructs encoding only miRNA binding sites or encoding multiple miRNA binding sites in tandem. In contrast, our assays used the 3′-UTR of endogenous genes to test APOBEC3G activity and we also tested how APOBEC3G blocks DND1 function on these 3′-UTRs. Moreover, the amount of APOBEC3G expression vectors transfected into cells were considerably higher in the study  and we also found that higher APOBEC3G levels may have alternate effects on the 3′-UTRs and miRNA function. Another possibility is that APOBEC3G may have different effects on different miRNAs and transcripts, inhibiting some miRNAs while activating others.
One caveat of our studies is that the miRNA studies were performed using human DND1 and APOBEC3G whereas the infectivity assays and genetic studies were performed using mouse factors. We did this because human DND1 function was previously characterized using human P27 3′-UTR and DND1 binding sites have been mapped in the 3′UTR of P27. Humans encode multiple members of the APOBEC3 family proteins (APOBEC3A, 3B, 3C, etc.) . We selected human APOBEC3G for our miRNA studies which is functionally most closely related to mouse APOBEC3. The role of the other human APOBEC3 factors in inhibiting DND1 activity remains to be determined.
APOBEC3 proteins are expressed in germ cells [13, 21, 49]. But although APOBEC3 proteins inhibit retrotransposition [17, 50, 51], to date, this has not been demonstrated in germ cells. In fact, Apobec3 null mice are normal and fertile . We found that double null Ter/Ter;Apobec3−/− mice have similar phenotype as Ter/Ter mice and lack of Apobec3 does not rescue the Ter/Ter (Dnd1−/−) phenotype to restore germ cells. This supports the idea that DND1 does not regulate the mutator activity of APOBEC3 in germ cells, and deregulated APOBEC3 mutator activity is not responsible for germ cell loss in Ter/Ter (Dnd1−/−) mice.
In summary, our studies focus on a novel aspect of APOBEC3 function in that we show APOBEC3G regulates DND1 function and in this way affects miRNA activity. Viral restriction and miRNA mediated gene silencing are evolutionarily related processes utilizing similar protein complexes, which localizes to cytoplasmic RNA granules. Our data provides compelling evidence that APOBEC3G may be involved in both these processes.