We have identified RCK/p54 as a protein that interacts with Ago2 in affinity-purified RISC assemblies to facilitate formation of cytoplasmic P-bodies and that acts as a general translational repressor in human cells. We have shown that depletion of RCK/p54 disrupts P-bodies and disperses the cytoplasmic localization of Ago2. Furthermore, depletion of RCK/p54 did not significantly affect the RNAi function of RISC, although general, miRNA-mediated, and
let-7-mediated translational repression were released. Together, our results provide significant insights into miRNA mechanisms in human cells (see below).
Several recent reports show that RISC co-localizes with P-bodies, suggesting that they could be the site where RISC degrades target mRNA or represses mRNA translation [
]. Isolation of miRNAs on polysomes also links the suppression of protein synthesis with arrest of translation initiation or elongation [
]. These studies also demonstrated that human Ago1 and Ago2 co-localize in P-bodies with other cellular proteins, such as Dcp1a, Dcp2, GW182, Lsm1, and Xrn1 [
]. A homolog of the P-body protein GW182 in
is the developmental timing regulator AIN-1, which also interacts with miRISCs and may target argonaute proteins to P-bodies [
]. To dissect and understand the relationship between RNAi function and P-bodies, we affinity-purified RISC using Myc-Ago2 and expression vectors of the YFP-tagged P-body proteins, Lsm1, RCK/p54, Dcp2, and eIF4E. Ago2 interacted with these various P-body components in ways that were RNA-dependent or RNA-independent (
A). Ago1 and RCK/p54 immunoprecipitated with Ago2 after RNase A treatment of HeLa cell extracts, suggesting that these proteins directly interact with Ago2. Interestingly, RCK/p54, Ago1, and Ago2 were also identified as a component of active RISC programmed with siRNA or miRNA and purified by biotin affinity to streptavidin-conjugated magnetic beads (
). We examined the P-body localization of Ago2 with Lsm1 and RCK/p54 by co-expressing YFP-tagged Lsm1 and RCK/p54 with CFP-Ago2. Interestingly, overexpressing YFP-RCK/p54 in HeLa cells increased the number of P-bodies (from ~ 8 to ~ 20 foci/cell). The number of P-bodies containing CFP-Ago2 also increased (
B). These results suggested a functional relationship between RCK/p54-Ago interactions and their localization to P-bodies. To visualize protein–protein interactions in vivo, we used FRET as a probe. In cells expressing YFP-RCK/p54 and CFP-Ago2, the FRET efficiency was 21.07% ± 2.52% (
D). We also observed an efficient FRET between YFP-Ago1 and CFP-Ago2; however, FRET between RCK/p54 and Ago1 was moderate (6.41% ± 1.96%). Since FRET is quite sensitive to the orientation of the donor: acceptor pair, it is possible that CFP and YFP in Ago1 and RCK/p54 are not suitably positioned for efficient energy transfer. Nonetheless, the FRET efficiency between Ago1 and RCK/p54 was significantly above the 0.9% background. Taken together, these results demonstrate that Ago1, Ago2, and RCK/p54 directly interact in vivo.
Where in the cell does this physical interaction between RISC and RCK/p54 take place and does it require P-body structures? To address these questions, we disrupted P-bodies in cells by depleting Lsm1 and immunopurified endogenous miRISC, and analyzed its ability to cleave a target mRNA with perfect complementarity to
let-7 miRNA (
). We observed no significant difference in miRISC activities purified from Lsm1-depleted and non-depleted cells (
) even though the P-body structures were disrupted and the total number of P-bodies per cell had significantly decreased. These results show that RCK/p54 association with RISC does not require P-bodies and suggest that localization of RISC to P-bodies is the outcome of translation repression by miRISC.
Having established the physical interactions of RCK/p54 with RISC, we questioned the functional relevance of these interactions in the RNAi pathway and how they could facilitate RISC activity. To that end, we depleted RCK/p54 in HeLa cells and analyzed RNAi activities in vivo and in vitro (
). Depletion of RCK/p54 clearly disrupted the P-body structures and dispersed Ago2 localization throughout the cytoplasm (
). Surprisingly, depleting RCK/p54 did not significantly affect siRNA-mediated gene silencing. Similar results showing the lack of P-body involvement in RNAi have been reported in
]. In that system, depletion of the P-body proteins, GW182 or Dcp1:Dcp2 decapping complex, showed that these proteins were required for miRNA-mediated gene silencing
but RNAi efficiency was not affected, suggesting that these two gene-silencing mechanisms were independent in
]. In addition, two recent studies reported that mammalian GW182 interacts with Ago2 and plays a role in miRNA function [
] and in siRNA-mediated RNAi [
]. Interestingly, depleting GW182 disrupted cytoplasmic foci and interfered with RNAi activity [
]. It is possible that GW182 is directly involved in human RISC function and does not spatially require P-bodies for its function. As an alternative approach to disrupting P-bodies, we treated HeLa cells with the translation inhibitor, cycloheximide, which did not significantly affect RNAi activities in vivo and in vitro (unpublished data). Taken together, these results strongly suggest that functional siRISC is assembled before locating to P-bodies, and P-bodies are not prerequisite sites for siRNA-mediated gene silencing.
To determine the role of RCK/p54 in miRNA-mediated translation repression, we used a CXCR4 reporter system [
], in which siRNA- or miRNA-reporter constructs harbor 1 × perfectly matched or 4 × bulged CXCR4 siRNA target sites, respectively in the 3′ UTR of RL mRNA [
]. In this system, perfectly matched sequences are cleaved via siRISC and bulge-containing sequences are targets for translation suppression by miRISC. We observed that depletion of RCK/p54 released only miRNA-mediated gene suppression and had no effect on siRNA-mediated gene silencing (
B). Depletion of GW182, an argonaute-interacting P-body protein, released gene suppression by miRNA and moderately affected reporter constructs containing perfectly matched target sequences, consistent with previous findings [
]. Ago2 depletion significantly released siRNA-mediated gene silencing and moderately affected reporters containing 4 × bulged sequences. Knockdown of Lsm1 had no significant effect on either miRNA- or siRNA-mediated gene silencing. These results show that RCK/p54 depletion releases miRNA-mediated translation suppression of reporter genes and that this function of RCK/p54 was not significantly affected when P-bodies were disrupted by depleting Lsm1.
To test our hypothesis that RCK/p54 mediates the translational repression of endogenous miRNA targets in P-bodies, we examined RAS protein levels in RCK/p54-depleted cells. We chose RAS because it is an endogenous target of
and 3′ UTRs of human
genes contain multiple complementary sites for
-7 to bind and regulate RAS expression levels [
-7 inhibitors are known to enhance RAS protein expression in HeLa cells [
]. Depleting RCK/p54 in HeLa cells up-regulated RAS protein, and this increase in RAS levels was higher than that in general translation of control actin (
C), suggesting that multiple sites of
-7 miRISC binding to target 3′ UTR dictate the potency and specificity of translation suppression.
RCK/p54, the human homolog of yeast Dhh1p, is a member of the ATP-dependent DEAD box helicase family and was originally identified as a proto-oncogene [
]. In human cells, RCK/p54 interacts in P-bodies with the translation initiation factor, eIF4E [
homolog of RCK/p54, Xp54, which interacts with eIF4E and forms RNA-dependent oligomers, represses the translation of mRNA in oocytes and eggs [
]. In yeast, Dhh1p interacts with the decapping and deadenylase complex and functions in translational repression [
]. Dhh1p has recently been shown to stimulate translational repression by inhibiting production of the pre-initiation complex [
]. Dhh1p and RCK/p54 also inhibit the mRNA translation driven by internal ribosomal entry sites in vitro, suggesting that this protein is a general translational repressor that may not require a cap structure for its function [
]. However, the in vivo function of Dhh1p required translational initiation [
]. Translation repression by miRISC and P-body localization require the 5′-cap structure in the target mRNA [
], which provides a unique and elegant control mechanism for translation and its regulation (reviewed in [
]). Recently, Petersen et al. reported an intriguing study showing that miRNA represses translation after initiation by a ribosome drop-off mechanism [
]. Therefore, it is possible that RCK/p54 interacts with miRISC and blocks translation elongation by binding to its multiple target sites with high affinity and creates a barrier for elongating ribosomes (see below).
We therefore propose that the function of RISC assemblies in cells can be described by a model with two independent pathways, depending on whether RNAi is programmed by the guide strand of siRNA or miRNA. In the first pathway, the RISC recognizes a perfectly matched target mRNA and functions as siRISC (
) by cleaving its target mRNA. This target mRNA cleavage activity and rapid turnover of siRISC induce the subsequent target mRNA degradation by exonucleases and does not require P-bodies. In the second pathway, the RISC recognizes the imperfectly matched target and functions as miRISC (
). Multiple copies of miRISC containing RCK/p54 could initiate an oligomerization event that sequesters the whole RNP and transports it to the P-bodies. This sequestration could also explain the cooperative effects of RISC function that enhance translation repression [
]. Once in P-bodies, translationally repressed mRNA could stay in oligomeric structures for storage or it could form a complex with decapping enzymes and cap binding proteins that would lead to the mRNA decay pathway. In other words, the miRNA in RISC could provide the sequence specificity and RCK/p54 could be the effector molecule that shuttles the target mRNA toward the fate of storage or processing in P-bodies. Since P-bodies are not a prerequisite for RCK/p54 to assemble into RISC or to function, it is possible that miRISC blocks protein synthesis by enhancing the ribosome drop-off during elongation of translation [
]. We propose that miRISC localization to P-bodies is the outcome of translation repression and not a prerequisite for miRISC function. Stored mRNA in P-bodies could reenter the active translation or mRNA decay pathway depending upon cellular conditions or stimuli that are presently unknown. How long mRNA translation can be repressed in the cytoplasm and which factors trigger mRNA transport to P-bodies are exciting areas for future research.
A Model for Human RISC Function Involving miRNA and siRNA
Our results on the interactions between RCK/p54 and the RNAi machinery suggest an intriguing role for miRNA function in development and carcinogenesis. However, most targets of miRNA have not yet been identified. A growing body of evidence suggests that miRNAs are important in human disease, including cancers [
]. For example, relatively low levels of
-7 miRNA up-regulate RAS protein in lung cancer cells, demonstrating a possible role of miRNA in tumorigenesis [
]. Two homologs of RCK/p54, Xp54 in
and Me31b in
control the translation of maternal mRNA in oocytes [
]. Altered regulation of RCK/p54 expression levels has been implicated in the development of human colorectal tumors [
] and in hepatitis C virus-related chronic hepatitis [
]. Overexpression of RCK/p54 and Dhh1p increases the number of Ago2-containing P-bodies (
]), suggesting that RCK/p54 is a critical factor in controlling cellular translation. Our results have identified RCK/p54 as an essential component of miRISC-mediated translation repression and provide a functional link between miRISC and P-bodies. In the context of spatially and temporally restricted miRNA expression, this system allows for exquisite regulation of protein expression by target mRNA. Therefore, perturbations of either miRNA or RCK/p54 expression levels can have deleterious consequences for the cell. What determines the balance between active translation and repression of mRNAs targeted by miRISC, and how cells control the specificity of this repression are key directions for future investigation.