Our data indicate that CCR4-NOT is the deadenylase that removes poly(A) from messages targeted by miRNAs in human cells, thereby triggering rapid mRNA degradation. Overproducing a mutationally inactivated form of either of the two catalytic subunits of this deadenylase (ccry or CAF1/POP2) significantly impedes the deadenylation of RISC-associated mRNAs, whereas overproducing catalytically inactive forms of the other two human deadenylases, PAN2-PAN3 and PARN, has no effect. These findings in human cells are consistent with the effect of depleting CAF1 or NOT1 in Drosophila
) and with recent independent reports based on studies in mouse fibroblasts and mouse ascites cell extracts (5
Human CCR4-NOT comprises nine subunits, two of which have intrinsic deadenylase activity (14
). Each of those catalytically active subunits exists in two paralogous forms in humans: CCR4a/CCR4b and CAF1/POP2. Overproducing inactive CAF1 or POP2 causes a markedly greater impediment to poly(A) removal than overproducing CCR4a, whereas excess production of the active forms of these proteins has no effect. These findings indicate that, of the two catalytic subunits in the CCR4-NOT complex, CAF1/POP2 makes a greater contribution to miRNA-mediated deadenylation than does CCR4. Furthermore, using RNA interference to reduce the cellular concentration of CAF1 significantly slows deadenylation, whereas knocking down POP2 synthesis to a similar extent does not, results that suggest a greater role for CAF1 than POP2 in poly(A) removal directed by miRNAs in HeLa cells. This difference may simply reflect a possible molar excess of CAF1 over POP2 in this cell type, since the failure of deadenylation to be impeded by wild-type POP2 overproduction argues against an intrinsic disparity in the ability of these two paralogs to degrade poly(A).
The regulatory effects of miRNAs are mediated by the protein components of RISC, which deliver miRNAs to the messages they target (8
). In principle, RISC could expedite poly(A) removal either by recruiting CCR4-NOT to mRNA, thereby increasing its concentration in the vicinity of the poly(A) tail, or by altering the poly(A) tail and its associated proteins in a way that increases the efficacy of deadenylation by CCR4-NOT without raising its local concentration. The former mechanism implies a physical association of RISC with CCR4-NOT, either direct or mediated by one or more other proteins specifically recruited by RISC. However, our coimmunoprecipitation data suggest that, notwithstanding a recent report to the contrary (9
), RISC does not associate significantly with CCR4-NOT or with either of the other two human deadenylases. Thus, despite the importance of the CAF1/POP2 subunit of CCR4-NOT for miRNA-mediated deadenylation, neither of these proteins was found to interact detectably with either Ago or TNRC6. Moreover, only trace amounts of CCR4a and CCR4b were observed to coimmunoprecipitate with Ago (corresponding to <1% of the level at which TNRC6 immunoprecipitated), and neither copurified detectably with TNRC6, the apparent effector subunit of RISC. It is possible that the weak association of CAF1 with Ago2 that was recently reported (9
) was an indirect consequence of bridging by the ubiquitous mRNP component PABP (40
). We conclude that RISC likely facilitates deadenylation by making the autonomous encounters of CCR4-NOT with poly(A) more productive. For example, the reported ability of TNRC6 to displace eIF4G from PABP (40
) may render poly(A) more susceptible to deadenylase attack. Such disruption of the eIF4G-mediated interaction of the cap-binding complex with PABP might also help to account for the inhibitory effect of miRNAs on translation, although it would not explain their ability to repress both cap-independent translation and the translation of mRNAs that lack a poly(A) tail as effectively as they repress the translation of capped, polyadenylated mRNAs (7
A complication of using RNA interference as a genetic tool is the propensity of transfected siRNAs to downregulate not only their intended targets but also many other mRNAs with which they can base pair with imperfect complementarity (3
). Our results indicate that these off-target effects are, at least in part, a consequence of mRNA destabilization resulting from poly(A) tail removal by CCR4-NOT. These findings support the view that synthetic siRNAs downregulate off-targets by the same mechanisms used by endogenous miRNAs to repress mRNAs to which they are partially complementary.
The multiplicity of functionally redundant Ago and TNRC subunits of RISC in mammalian cells has made tethering a valuable strategy for examining the respective contributions of those proteins and their component domains to gene silencing by miRNAs. Previously, we have shown that tethering any of the four human Ago proteins to mRNA, presumably as a complex with TNRC6, reduces both the concentration and translation efficiency (protein synthesis per message) of that mRNA (35
). Our present data show that this effect on mRNA concentration is a consequence of expedited mRNA decay triggered by rapid deadenylation. Moreover, deadenylation mediated by tethered Ago involves CCR4-NOT, the same deadenylase whose action is stimulated by miRNAs. Thus, in every respect, the effects of recruiting RISC to mRNA by tethering its Ago subunit mimic those of recruiting RISC by miRNA base pairing to an imperfectly complementary mRNA element. These findings confirm the validity of RISC tethering as a surrogate methodology for investigating miRNA-mediated silencing.