This study shows that expression of a putative RNA/DNA editing enzyme, APOBEC2, is regulated by TGFβ signaling, and in turn regulates the pathway by ligand-specific inhibition, generating a negative feedback loop. This work also assigns a previously unrecognized biological function for APOBEC2 in the context of axis determination during early embryogenesis. The contribution of APOBEC2 to the TGFβ pathway, its mechanism of action, and its role during left-right specification are all unexpected, and discussed below.
We show that APOBEC2 (A2) regulates TGFβ signaling. A2 is a member of the cytidine deaminase family of DNA/RNA editing enzymes that emerged at the beginning of vertebrate radiation (Conticello, 2008
). These enzymes can mutate C to U in DNA or RNA (Conticello et al., 2007
; Navaratnam and Sarwar, 2006
; Smith, 2008
). The other members of the family are APOBEC1, AID and APOBEC3s (with several variants in primates). AID and APOBEC2 are the only members conserved among chordates, while APOBEC1 and APOBEC3s are mammalian-specific (Conticello et al., 2007
). APOBEC1 edits apolipoprotein B mRNA in a tissue specific manner by introducing a premature stop codon leading to a shorter protein. AID has been shown to edit the immunoglobulin locus directly on genomic DNA to generate antibody diversification. Different variants of APOBEC3 protein edit retroelements and retroviruses, such as HIV-1, to provide retroviral immunity. The fact that TGFβ signaling regulates expression of xA2 suggests that, just as in the context of immuno-diversification or retroviral editing, non-autonomous signals control editing activity in early embryogenesis. However, while editing functions described for previous members of the family affect cellular activities without changing their fate, A2 affects cell fate decisions with implications in axis development. Selective inhibition of Derrière, but not Xnr1, by A2 also provides a hint to how signaling by similar TGFβ ligands using the same receptor complex and signal transducers (Smad2 and 3 for nodal and derrière) can be segregated by autonomous factors. A2 provides a negative feedback loop only for derrière, but not for Xnr1, presumably when presented as homodimers. The fate of signaling by heterodimers of the two ligands has not been addressed by this study.
Interestingly, it was recently shown that mA2 is required for normal muscle development. mA2-deficient mice exhibit slightly reduced muscle mass, increased slow:fast fiber ratio, and developed myopathy at 8-10 months of age (Sato et al., 2009
). We show here that A2 is required for terminal differentiation of C2C12 myoblasts in vitro
and that it can inhibit TGFβ signaling. It is tempting to speculate that the defects observed in vivo
in mutant mice might be mediated by increased TGFβ signaling due to GDF-type ligands expressed in muscle cells, such as Myostatin/GDF8 and GDF11 (Artaza et al., 2002
; Lee et al., 2005
), in a A2-deficient background. This is also supported by the decrease in slow:fast fiber ratio seen in myostatin-deficient mice (Amthor et al., 2007
; Girgenrath et al., 2005
), indicating that APOBEC2 and myostatin have opposite effects on muscle in vivo
Regarding the mechanism of the A2 effect in the context of left-right specification, it can act either as a DNA/RNA editing enzyme or by an independent, currently unknown, mechanism. While our experiments do not address this directly, the fact that conserved amino acids in the enzymatic domain required for the function of well established cytidine deaminases are also required for A2 biological activity suggests that A2 could at an editing level. Alternatively, A2 could, directly or indirectly, affect the translation or stability of proteins involved in TGFβ signaling.
If A2 acts as an editing enzyme, does it edit DNA, RNA, or both? The answer to this question is currently unknown. At the genomic level, APOBEC2 has been suggested to provide global changes in the methylation state of DNA in zebrafish embryos (Rai et al., 2008
). While our study does not address this directly, a number of observations suggest that amphibian A2 is not involved in global changes of genomic DNA. First, the ligand-specific nature of the TGFβ inhibitory effect, directed at Derrière but not nodal, is not consistent with global methylation changes. Second, we note that overexpression of zAPOBEC2a (one of the two variants in the fish), unlike zAID and zAPOBEC2b, had no effect on genomic demethylation (Rai et al., 2008
). This also suggests that the amphibian A2 is more similar to APOBEC2a, rather than the APOBEC2b ortholog in the fish. Finally, while in the fish both AID and APOBEC2 have been implicated in demethylation, we find that only A2 and not AID displays inhibitory activity in C2C12 cells, thus arguing against demethylation being causal to the activity.
We show that A2 activity is necessary for the specification of the left-right axis downstream of signaling. In agreement with this role, xA2
expression overlaps that of derrière
in cells of the posterior mesoderm. Evidence for the appropriate timing of A2 activity in the context of left-right specification is provided by our rescue experiments where the laterality defects generated by xA2 depletion were rescued by induction of Smad2 activity at early neurula stages. This is the same developmental stage when the posterior mesoderm acts as a signaling center to pattern the left lateral late mesoderm, through Derrière and Xnr1 signaling. We therefore propose that selective inhibition of Derrière signaling in posterior mesoderm by APOBEC2 is required for the process underlying specification of laterality in the left lateral plate mesoderm. While this explains why A2 is necessary for proper left-right axis specification, and links the phenotypic outcome to molecular events, it also raises interesting questions. First, neither MO-mediated reduction of A2 function in zebrafish (Rai et al., 2008
), nor knockout of APOBEC2 function by insertional disruption in the mouse (Mikl et al., 2005
) have been reported to affect laterality. In the mouse, as stated above, other cytidine deaminases, (for example APOBEC3s), are present, that can theoretically compensate redundantly for lack of APOBEC2. In the zebrafish study, depletion was achieved with the same morpholino oligonucleotide as the one used in our work, and a possible explanation would be that a discreet phenotype like heart orientation could have escaped observation. Second, the mechanism of the selective effect of A2 on Derrière/GDF, as opposed to nodal signaling, remains unexplained. It remains to be seen if other differences reported on the molecular pathways activated by Nodal versus Derrière (Dorey and Hill, 2006
; Howell et al., 2002
) are connected to the role of A2.
Regardless of the outcome, A2, and perhaps other members of the DNA/RNA editing family of enzymes, provides yet another unexpected level of regulation for the TGFβ signaling pathway.