PAZ PIWI domain (PPD) proteins, together with the RNA cleavage products of Dicer, form ribonucleoprotein complexes called RNA‐induced silencing complexes (RISCs). RISCs mediate gene silencing through targeted messenger RNA cleavage and translational suppression. The PAZ domains of PPD and Dicer proteins were originally thought to mediate binding between PPD proteins and Dicer, although no evidence exists to support this theory. Here we show that PAZ domains are not required for PPD protein–Dicer interactions. Rather, a subregion of the PIWI domain in PPD proteins, the PIWI‐box, binds directly to the Dicer RNase III domain. Stable binding between PPD proteins and Dicer was dependent on the activity of Hsp90. Unexpectedly, binding of PPD proteins to Dicer inhibits the RNase activity of this enzyme in vitro. Lastly, we show that PPD proteins and Dicer are present in soluble and membrane‐associated fractions, indicating that interactions between these two types of proteins may occur in multiple compartments.
The RNA interference (RNAi) machinery is used by eukaryotes to inhibit gene expression by sequence‐specific cleavage of messenger RNA (mRNA) or translational repression (Hannon, 2002; McManus & Sharp, 2002; Schwarz & Zamore, 2002). The specificity of these processes is dependent on 21–22‐nt small interfering RNAs (siRNAs) or microRNAs (miRNAs) produced from double‐stranded (ds) RNAs or hairpin precursors, respectively. In some eukaryotes, the RNAi apparatus also participates in the control of chromatin structure, although it is not clear how this process occurs (Schramke & Allshire, 2003, and references therein).
Two classes of proteins are central to RNAi: members of the PAZ PIWI domain (PPD) and Dicer families. PPD proteins are required at the effector stages of RNAi (Hammond et al, 2001) and in some cases for maturation of miRNAs (Grishok et al, 2001). Family members are defined by the presence of a central PAZ domain and a carboxy terminal PIWI domain (Carmell et al, 2002). Perhaps the best‐characterized proteins in RNAi pathways are the Dicer enzymes, which are dsRNA‐specific RNases that cleave dsRNA into 21–25‐nt base‐paired fragments (Bernstein et al, 2001; Billy et al, 2001; Ketting et al, 2001). Dicer cleavage products are incorporated into PPD‐containing RNA‐induced silencing complexes (RISCs), which presumably mediate mRNA degradation, translational suppression or chromatin silencing (Hannon, 2002; McManus & Sharp, 2002).
Engagement of Dicer with PPD proteins is probably required for the transfer of siRNAs/miRNAs to RISCs (Zamore, 2002). Indeed, physical interactions between PPD proteins and Dicer have been demonstrated (Hammond et al, 2001; Sasaki et al, 2003), but such interactions have not been well characterized. The situation appears to be more complicated in some invertebrate metazoans in which dsRNA‐binding proteins such as RDE‐4 or R2D2 are thought to link Dicer enzymes to RISCs (Tabara et al, 1999; Parrish & Fire, 2001; Liu et al, 2003). Interestingly, homologues of RDE‐4 and R2D2 have not been identified in humans or other mammalian species, and thus transfer of siRNAs/miRNAs to mammalian RISCs may use different mechanisms. Initially, heterotypic PAZ–PAZ interactions were thought to mediate the formation of PPD–Dicer complexes, but this seems unlikely now because a number of Dicer isoforms that do not have PAZ domains have been described (Schauer et al, 2002). In addition, a fragment of the human PPD protein EIF2C1 (AGO1) that contained the PIWI domain interacted more strongly with Dicer than a region containing only the PAZ domain (Doi et al, 2003).
In this study, we have characterized in detail the interactions between Dicer and two distantly related PPD proteins, Hiwi and AGO2. We report that direct interactions between PPD proteins and Dicer do not require PAZ domains, but are rather mediated through their PIWI and RNase III domains. Surprisingly, the binding of PPD protein to Dicer inhibits RNase activity in vitro. Finally, we found that interactions between PPD and Dicer proteins are dependent on Hsp90 activity.
Results and Discussion
Mapping of interaction domains
The nature of PPD–Dicer interactions, which is probably important for RNAi, is poorly understood. As a first step towards a more thorough understanding of this process, we mapped the regions within PPD proteins and Dicer that mediate such interactions. Two distantly related human PPD proteins, AGO2 (also known as GERp95) (Cikaluk et al, 1999) and Hiwi (Sharma et al, 2001), were chosen for these studies (Fig 1A). Along their entire lengths, the level of identity between these two proteins is approximately 22.7%.
To establish whether either of the two PPD signature domains is sufficient for binding to Dicer, the PAZ and PIWI domains of AGO2 and Hiwi were co‐expressed as GST‐fusions together with CFP–Dicer in HEK293T cells followed by isolation on glutathione–sepharose beads. Bound fractions were analysed by SDS–PAGE and immunoblotting with antibodies (Abs) to GST or GFP (Fig 1B). Stable interactions between CFP–Dicer and the AGO2 or Hiwi PAZ domains were not detected; however, the PIWI domains of AGO2 and Hiwi efficiently bound to CFP–Dicer (Fig 1B). During the preparation of this work, a study was published showing that a region of AGO1 that includes the PIWI domain bound to Dicer (Doi et al, 2003). However, this study did not address whether this was the case for other PPD proteins nor did it identify the region of Dicer that binds to AGO1.
To determine which Dicer domain interacts with PPD proteins, plasmids encoding GST fused to Dicer or its subdomains (Fig 2A) were transfected into HEK293T cells stably overexpressing AGO2, and lysates were prepared and subjected to glutathione–sepharose pulldown assays. These experiments clearly show that the PAZ domain of Dicer is not required for binding to AGO2 (Fig 2B). Rather, AGO2 bound only to full‐length Dicer or its carboxy terminal region, a region that lacks the PAZ domain but contains the dsRNA‐binding domain (DSRM) and RNase III domains (Fig 2A). Together, these results indicate that PAZ domains are not involved in interactions between PPD and Dicer proteins.
PPD–Dicer interactions are direct
We investigated whether associations between Dicer and PPD proteins are direct by first analysing interactions between purified proteins. Human Dicer was expressed in insect cells and purified as described (Zhang et al, 2002). GST–PPD fusion proteins were expressed in COS1 cells and purified on glutathione–sepharose beads (Tahbaz et al, 2001). Protein A–sepharose beads coated with anti‐Dicer Ab (Billy et al, 2001) or as a control, with Ab against human GAR1, were used for the immunoprecipitation of complexes of Dicer with GST–AGO2, GST–Hiwi or GST alone (Fig 3A). Proteins co‐immunoprecipitating with Dicer were visualized by immunoblotting, using anti‐GST Ab. GST–AGO2 and GST–Hiwi were retained by anti‐Dicer beads but not control anti‐GAR1 beads (Fig 3A, upper and middle panels) or noncoated Protein A beads (not shown). Neither anti‐Dicer nor anti‐GAR1 beads retained GST alone (Fig 3A, lower panel). To rule out the possibility that double‐strand or single‐strand RNAs are required for stable binding of Dicer to PPD proteins, nuclease (micrococcal nuclease or a cocktail of RNases V1 and A) digestion steps were included before pulldowns and immunoblot analyses. The nuclease treatments had no measurable effects on the stability of the Dicer–PPD protein complexes (data not shown). Together, these data suggest that interactions between Dicer and PPD proteins are direct.
To eliminate the possibility that proteins co‐purifying with either Dicer or GST–PPD preparations are involved in binding, we further investigated Dicer–PPD protein interactions using the yeast two‐hybrid assay. Saccharomyces cerevisiae does not encode PPD or Dicer homologues and therefore, if interactions between PPD proteins and Dicer are detected in this system, it is likely that they result from direct binding. Plasmids encoding the DSRM, and the RNase III domain A or B of Dicer (see Fig 2A) fused to the activation domain of Gal‐4, and plasmids encoding AGO2 and Hiwi PAZ, PIWI and PIWI‐box regions (see Fig 1A) fused in‐frame to the DNA‐binding domain of Gal‐4 were constructed. The PIWI‐box is a highly conserved motif present in all PIWI domains (Cox et al, 1998). Using this method, no interactions between PAZ domains of AGO2 or Hiwi and Dicer were evident (see supplementary table 2 online). In contrast, interactions between the PIWI domains of AGO2 and Hiwi and the RNase III‐A domain of Dicer were detected, although the interactions were not particularly strong. The strongest interactions occurred between the RNase III‐A domain of Dicer and the PIWI‐boxes of AGO2 and Hiwi (Fig 3B; see supplementary table 2 online). Relatively weak interactions were observed between the PIWI‐boxes and RNase III‐B. Interactions were not detected between PPD protein domains and the DSRM of Dicer (see supplementary table 2 online). Together, these results indicate that the PIWI‐box of PPD proteins interacts directly with the RNase III domains of Dicer.
Dicer activity is inhibited by PPD proteins in vitro
Dicer cleavage products (siRNAs/miRNAs) are incorporated into PPD‐protein‐containing RISCs that are the effectors in RNAi or translation inhibition. Presumably, interactions between PPD proteins and Dicer are required for the transfer of siRNAs or miRNAs from Dicer to RISCs (Baulcombe, 2001). Indeed, previous studies have documented the interaction between PPD proteins and Dicer (Hammond et al, 2001; Doi et al, 2003), but the effects of these interactions on Dicer activity have not been investigated. Given that PPD proteins bind to the RNase III domain of Dicer, we reasoned that this interaction could affect Dicer activity. To test this possibility, the activity of purified Dicer was assayed in the presence of GST fused to AGO2 or Hiwi. A dose‐dependent decrease in Dicer activity was observed in the presence of GST–AGO2 and GST–Hiwi but not of GST alone (Fig 4). Similar results were obtained when Dicer and AGO2 were incubated together before the addition of the dsRNA substrate or when the pre‐let7 RNA hairpin was used as a processing substrate instead of dsRNA (data not shown).
Hsp90 activity is required for PPD–Dicer interaction
Previously, we reported that the AGO2 orthologue GERp95 interacts with an Hsp90‐containing protein complex (Tahbaz et al, 2001). These interactions are thought to be important for the biogenesis of GERp95 because inhibition of Hsp90 activity with geldanamycin (GD) or radicicol resulted in the preferential degradation of nascent GERp95 and the depletion of membrane‐bound pools of this protein. GD and radicicol bind specifically and stably to the ATP‐binding site of Hsp90 (Prodromou et al, 1997; Stebbins et al, 1997; Schulte et al, 1998), thereby disrupting the chaperone cycle. Accordingly, these inhibitors negatively affect the activities and localizations of many Hsp90 substrates. To determine whether Hsp90 activity is important for the formation and/or stability of PPD protein–Dicer complexes, we examined the effect of GD on Dicer interactions with AGO2 and Hiwi. The association of CFP–Dicer with GST–AGO2 (Fig 5A), and GST–Hiwi (Fig 5B) was greatly inhibited by GD, indicating that Hsp90 has a role in regulating the binding of multiple PPD proteins to Dicer. The GD‐dependent inhibition of Dicer–PPD interactions was confirmed by reciprocal pulldown assays using GST–Dicer expressed transiently in stable cell lines overexpressing AGO2 (data not shown). It should be noted that GD treatment caused some degradation of PPD proteins; however, during the short incubation times used here, the extent of GST–AGO2 and GST–Hiwi degradation was not sufficient to account for the near‐complete GD‐dependent block of Dicer binding.
Two cellular pools of AGO2, Hiwi and Dicer
Our previous work showed that the rat orthologue of AGO2 is present in membrane‐associated and cytosolic pools (Cikaluk et al, 1999; Tahbaz et al, 2001). Recent studies have shown that Dicer also localizes to the cytoplasm (Billy et al, 2001; Provost et al, 2002; Findley et al, 2003), but quantitative evidence of membrane association has not been reported. The intracellular site(s) of PPD–Dicer interactions have not been investigated, and therefore we compared the intracellular distributions of PPD proteins and Dicer in subcellular fractions. Similar to AGO2, a large pool of Dicer was present in soluble fractions, whereas a smaller but significant cohort of this RNase co‐purified with membranes (Fig 6). Calnexin, an integral membrane protein of the endoplasmic reticulum, was used as a marker for the membrane fraction. Hsp90 was present only in the soluble fractions. As antibodies to Hiwi are not available, the distribution of GST–Hiwi expressed in HEK293T cells was examined. Similar to AGO2 and Dicer, GST–Hiwi was present in both soluble and membrane‐associated fractions (Fig 6).
Conclusions and Speculation
Our data indicate for the first time that interactions between Dicer and PPD proteins are direct, and that the Dicer RNase III domain is required for binding to PPD proteins. Moreover, the results from this study point towards the existence of stable Dicer–PPD complexes in multiple locations (cytosol and membranes). The existence of two pools of Dicer–PPD complexes may reflect the roles of these proteins in mRNA degradation or translation taking place on free and membrane‐associated polysomes. Indeed, association of RISC complexes and siRNAs with ribosomes has been previously documented (Hammond et al, 2001; Djikeng et al, 2003). Likewise, a fraction of Dicer present in mouse P19 cell extracts sediments in the 80S region of gradients (our unpublished results), but the relevance of Dicer association with ribosomes is unclear.
Interactions between Dicer and PPD proteins (Hammond et al, 2001; Doi et al, 2003; Sasaki et al, 2003, and this work) are likely to be important for the transfer of siRNAs generated by Dicer to RISC complexes, of which PPD proteins are a part (Hammond et al, 2001; Hutvagner & Zamore, 2002; Martinez et al, 2002). Notably, by interacting with the RNase III domain of Dicer, as demonstrated here, PPD proteins are brought to the proximity of siRNAs produced by the catalytic domain of the enzyme. In this context, it is important to note that, in reactions carried out with the purified recombinant Dicer, a considerable fraction of siRNA products was found to remain associated with the enzyme (Zhang et al, 2002). Hence, the interaction with PPD proteins or RISC complexes might be expected to stimulate the activity of Dicer by facilitating the product release. However, rather unexpectedly, we found that PPD proteins inhibit the activity of the purified Dicer in vitro. It is possible that, physiologically, active Dicer molecules are not bound to PPD proteins or that Dicer–PPD protein complexes contain other components that activate Dicer or modulate the inhibitory role of PPD proteins. If the former is true, Hsp90 is a candidate that could indirectly influence Dicer activity by regulating the binding of PPD proteins.
In summary, the likely existence of two pools of Dicer–PPD complexes, together with a role for Hsp90 in PPD–Dicer associations and the inhibitory effect of PPD proteins on Dicer activity in vitro, indicates that Dicer and PPD proteins are parts of a rather complex and dynamic network of interactions important for executing siRNA and miRNA functions.
Materials and Methods
Materials. The list of materials and suppliers can be found in the supplementary information online.
Yeast two‐hybrid assay. Coding regions for human Dicer or PPD protein (AGO2 and Hiwi) subdomains were fused with the Gal‐4 activation domain of pGADT7 and the Gal‐4 DNA‐binding domain of pGBKT7, respectively (see supplementary information online). The resulting plasmid constructs were transformed into S. cerevisiae strain AH109 and plated on complete minimum media (CMM) lacking leucine and tryptophan for 2–3 days at 30°C. Transformants were then streaked onto media lacking leucine, tryptophan, histidine and adenine. In cases where colonies grew in less than 7 days, these combinations were deemed as evidence of positive interactions, whereas lack of growth after 14 days was deemed as a negative interaction.
In vitro Dicer activity assays. DsRNA processing assays were performed as previously described (Zhang et al, 2002), and contained 130‐bp dsRNA uniformly labelled with [α‐32P]UTP, 20 ng of the recombinant human Dicer–HisC protein, and indicated amounts of GST–Hiwi, GST–AGO2 or GST.
Other procedures. Detailed methods including plasmid construction, GST‐pulldowns and immunoprecipitations can be found in the supplementary information online.
Supplementary information is available at EMBO reports online (http://www.nature.com/embor/journal/vaop/ncurrent/extref/7400070‐s1.pdf).
We thank Dr Henry Parker for help with subcloning and Margaret Hughes for cell culture support. This work was funded by a grant from the Canadian Institutes of Health Research (CIHR) (T.C.H). T.C.H. is the recipient of a Senior Medical Scholarship from the Alberta Heritage Foundation for Medical Research (AHFMR). N.T. and K.J. are supported by pre‐doctoral studentship awards from CIHR and AHFMR, respectively. F.A.K. was the recipient of a long‐term fellowship from the Human Frontier Science Program. Friedrich Miescher Institute is a part of the Novartis Research Foundation.
- Copyright © 2004 European Molecular Biology Organization