|Home | About | Journals | Submit | Contact Us | Français|
Emerging evidence suggests that RNAi-related processes act both in the cytoplasm and in the nucleus. However, the process by which the RNAi machinery is transported into the nucleus remains poorly understood. The Tetrahymena Argonaute protein Twi1p localizes to the nucleus and is crucial for small RNA-directed programmed DNA elimination. In this study, we identify Giw1p, which binds to Twi1p and is required for its nuclear localization. Further, the endoribonuclease (Slicer)-activity of Twi1p plays a vital role in the removal of one of the two strands of Twi1p-associated siRNAs, leading to a functionally mature Twi1p-siRNA complex. Slicer activity is also shown to be required for nuclear localization of Twi1p and for its association with Giw1p. These results suggest that Giw1p senses the state of Twi1p-associated siRNAs and selectively transports the mature Twi1p-siRNA complex into the nucleus.
Argonaute family proteins bind to small RNAs (~20-30 nt) and are integral players in all known RNAi-related gene regulatory pathways (reviewed in Tolia and Joshua-Tor 2007). Many Argonaute proteins act in the cytoplasm, where they induce post-transcriptional gene silencing. Recent evidence suggests that Argonaute proteins also act in the nucleus.
In mammals, the Argonaute proteins Ago1 and Ago2 mediate transcriptional silencing (Janowski et al. 2006, Kim et al. 2006) and Ago2 localizes to the nucleus in an Importin 8-dependent manner (Weinmann et al. 2009). Another Argonaute protein, MIWI2, localizes to the nucleus in fetal mouse testes and is required for DNA methylation-mediated retrotransposon silencing (Aravin et al. 2008, Kuramochi-Miyagawa et al. 2008). The Drosophila Argonaute protein Piwi localizes to nuclei of nurse and follicle cells in the ovary (Cox et al. 2000, Brennecke et al. 2007) and plays a role in transcriptional gene silencing (Pal-Bhadra et al. 2002). In Arabidopsis, the nuclear-localizing Argonaute proteins AGO4 and AGO6 are involved in RNA-directed DNA methylation (Li et al. 2006, Pontes et al., 2006, Zheng et al. 2007). In the fission yeast Schizosaccharomyces pombe, the Argonaute protein Ago1 is involved in both transcriptional and post-transcriptional gene silencing (Volpe et al., 2002, Sigova et al., 2004) and localizes to both cytoplasm and nucleus (Noma et al., 2004).
Two recent studies indicate that nuclear import of some Argonaute proteins is dependent on small RNAs. The Caenorhabditis elegans Argonaute protein NRDE-3 needs to associate with a siRNA to localize to the nucleus (Guang et al. 2008). In mice, nuclear localization of MIWI2 requires MILI, which is essential for the production of MIWI2-associated piRNAs (Aravin et al. 2008). These studies suggest that some mechanism distinguishes between free Argonaute proteins and those complexed with small RNAs, transporting only the latter into the nucleus. However, little is known about how small RNAs regulate the nuclear localization of Argonaute proteins.
The ciliated protozoan Tetrahymena thermophila provides an extreme example of a nuclear-acting Argonaute protein. The Argonaute protein Twi1p plays an essential role in programmed DNA elimination (Mochizuki et al. 2002), which is evolutionarily related to RNAi-directed heterochromatin formation in other eukaryotes (reviewed in Malone and Hannon 2009). Tetrahymena possesses a germline micronucleus and a somatic macronucleus in a single cell. The micronucleus produces both new micro- and new macronuclei during sexual reproduction. During macronuclear development, ~6000 different internal eliminated sequences (IESs) are defined by ~28-29 nt siRNAs, termed scnRNAs (Mochizuki et al. 2002, Yao et al. 2003, Lee and Collins 2006), and removed. scnRNAs are processed from bi-directionally transcribed noncoding RNAs (Chalker and Yao 2001) by the Dicer-like protein Dcl1p in the micronucleus (Malone et al. 2005, Mochizuki and Gorovsky 2005) and complex with Twi1p in the cytoplasm (Mochizuki and Gorovsky 2004). The Twi1p-scnRNA complex is then transported into the parental macronucleus where it has been proposed that IES-specific scnRNAs are enriched through selective degradation of scnRNAs that are complementary to the macronuclear DNA probably by recognizing nascent transcripts (Mochizuki et al. 2002, Aronica el al. 2008). Finally, Twi1p-scnRNA complexes move into the developing macronucleus, where they induce the formation of heterochromatin, leading to DNA elimination (Mochizuki et al. 2002, Taverna et al. 2002, Liu et al. 2007).
These dynamic changes in the localization of Twi1p are believed to be essential for the conserved small RNA-directed heterochromatin formation process, providing an attractive model to study how Argonaute-small RNA complexes are localized and how their localizations influence their functions in eukaryotes. Here we report that nuclear localization of Twi1p is regulated by the Twi1p-binding protein Giw1p, which senses the state of siRNAs associated with Twi1p. This mechanism enables Tetrahymena to transport only a functionally mature Argonaute-siRNA complex into the nucleus.
Some Argonaute proteins have endoribonuclease (Slicer) activity responsible for cutting RNAs with sequences complementary to those of their small RNA cargos. Slicer activity is provided by the evolutionarily conserved Piwi domain, when it contains a conserved catalytic core composed of an Asp-Asp-His (DDH) motif (reviewed in Tolia and Joshua-Tor 2007). Comparison of the Piwi domains of the Tetrahymena Argonaute protein Twi1p and other Argonaute proteins possessing Slicer activity (Fig. S1A, Couvillion et al. 2009) revealed that Twi1p contains a DDH motif (Asp526-Asp596-His745), suggesting that it may have Slicer activity.
Slicer activity of Twi1p was analyzed using recombinant Twi1p expressed in E. coli as a GST fusion protein. GST-Twi1p was incubated with 27 nt “guide” RNA to form RISC-like ribonucleoprotein complexes. These complexes were then incubated with a 5′ end radio-labeled 27 nt “substrate” RNA whose 3′ 25 nt were complementary to the guide RNA. As a positive control, recombinant human Ago2 fused to GST (Rivas et al. 2005) was prepared, complexed with 21 nt guide RNA and incubated with a 5′ end radio-labeled 21 nt substrate RNA whose 3′ 19 nt were complementary to the guide RNA. The cleaved product was observed by denaturing gel-electrophoresis followed by autoradiography (Fig.1A). If the GST-Twi1p cleaves the substrate in a similar manner to other Argonaute proteins with Slicer activity which cleave the bond between residues base-paired to nucleotides 10 and 11 of the guide strand (reviewed in Tolia and Joshua-Tor 2007), the radio-labeled cleavage product should be 15 nt long (see schematic drawing in Fig. 1A). Indeed, a 15 nt RNA species was detected in the GST-Twi1p sample (Fig. 1A, 27 mer, WT), but not when using GST alone (Fig. 1A, 27 mer, GST). Similarly, radiolabeled cleavage product indicating cutting between residues base-paired to nucleotides 10 and 11 of the guide strand (= 9 nt), was detected using GST-hAgo2 and 21 nt RNAs (Fig. 1A, 21 mer, GST-hAgo2). We conclude that Twi1p possesses Slicer activity.
Cleavage of substrate RNA by GST-Twi1p in our assay was less efficient than that by some other Argonautes in similar assays (Matranga et al. 2005, Miyoshi et al. 2005, Rand et al. 2005). A pre-steady state kinetics analysis of substrate RNA cleavage (Förstemann et al. 2007) suggested that only ~0.4 % of the recombinant GST-Twi1p was active (Fig. S1B). The inefficient substrate cleavage by GST-Twi1p may be caused either by enzymatic inactivity of the majority of GST-Twi1p or by inefficient complex formation between Twi1p-guide-strand scnRNA in vitro.
To determine if the DDH motif is involved in the Slicer activity of Twi1p, mutants were created in which either the first aspartic acid of the motif was replaced by asparagine (Twi1p-D526N) or the last histidine of the motif was replaced by glutamine (Twi1p-H754Q) (Fig. S1A). Twi1p-D526N or Twi1p-H745Q fused to GST was analyzed as above. As shown in Fig. 1A, neither mutated enzyme produced a detectable 15 nt cleavage product, suggesting that the conserved motif is required for the Slicer activity of Twi1p. These mutated Twi1p proteins are referred to as Slicer-dead Twi1p.
In vivo, Twi1p associates with ~28-29 nt siRNAs, named scnRNAs (Mochizuki et al 2004). Since scnRNAs are processed from double-stranded noncoding RNA by the Dicer-like protein Dcl1p (Malone et al. 2005, Mochizuki and Gorovsky 2005), there must be a mechanism responsible for making scnRNAs single-stranded. Because the Slicer activities of several Argonaute proteins are involved in the removal of one of the two small RNA strands (passenger strand) in other organisms (Matranga et al. 2005, Miyoshi et al. 2005, Rand et al. 2005, Leuschner et al. 2006, Maiti et al. 2007, Steiner et al. 2009), we tested whether the Slicer activity of Twi1p was involved in scnRNA passenger strand removal in vivo.
We constructed Tetrahymena strains whose TWI1 loci in the polyploid macronuclus were completely replaced by a mutant construct (FLAG-HA-TWI1-D526N) encoding Slicer-dead Twi1p-D526N tagged with FLAG-HA (Fig. S1B, S1D). FLAG-HA-TWI1-WT, expressing wild type Twi1p tagged with FLAG-HA (Fig. S1B, S1C), was also used. Two FLAG-HA-TWI1-WT strains of different mating types produced viable sexual progeny (Fig. S1M), indicating that the FLAG-HA tag does not disturb the essential function (Mochizuki et al. 2002) of Twi1p. Comparable amounts of FLAG-HA-Twi1p-WT and FLAG-HA-Twi1p-D526N were detected by western blotting (Fig. S1G), indicating that the Slicer activity is not required for the accumulation of Twi1p protein.
scnRNAs co-immunoprecipitated with FLAG-HA-Twi1p-D526N or FLAG-HA-Twi1p-WT from cells at an early stage (4 hr post-mixing) of conjugation were separated in native gel and stained with GelRed (Fig. 1B). More than half of the FLAG-HA-Twi1p-D526N-associated scnRNAs detected by the staining migrated at positions corresponding to double-stranded ~28 nt RNA (Fig. 1B, asterisk). The remaining scnRNAs from FLAG-HA-Twi1p-D526N and all scnRNAs from FLAG-HA-Twi1p-WT migrate as a smear (Fig. 1B, bracket). We believe this smear signal is attributable to the extensive sequence heterogeneity of the scnRNAs, which are believed to be transcribed from the whole micronuclear genome (Mochizuki et al 2002). Consistent with this, by northern hybridization, two different 28 nt oligo DNA probes complementary to different specific scnRNAs sequences detected distinct bands within the smear region (Fig. S1N). Because of their small size and AT richness, fraction of scnRNAs could dissociate during experimental handling and this analysis likely underestimates the amount of double-stranded scnRNA associated with FLAG-HA-Twi1p-D526N. In contrast, none of the scnRNA associated with FLAG-HA-Twi1p-WT migrated to the position on the gel corresponding to double-stranded scnRNA (Fig. 1B, WT). Denaturing gel analysis of these scnRNAs indicated that similar amounts of scnRNA were associated with FLAG-HA-Twi1p-D526N and FLAG-HA-Twi1p-WT at 2 and 4 hr post-mixing (Fig. 1C). We conclude that the Slicer activity of Twi1p plays an important, possibly essential, role in the scnRNA passenger strand removal in vivo.
We analyzed the expression of scnRNAs in the absence of the Slicer activity of Twi1p. FLAG-HA-TWI1-D526N strains expressed similar levels of scnRNAs to FLAG-HA-TWI1-WT strains in the early stages of conjugation (Fig. 1D, 2-4 hr post-mixing). However, in the FLAG-HA-TWI1-D526N strains, the amount of scnRNA was greatly reduced at the mid stage of conjugation (Fig. 1D, 6 hr) and became undetectable at later stages of conjugation (Fig. 1D, 8-10 hr). Thus, Slicer activity of Twi1p is not required for production of scnRNAs, but is required for their stable accumulation. An exonuclease likely degrades the double-stranded scnRNAs complexed with Slicer-dead Twi1p, as the scnRNAs associated with FLAG-HA-Twi1p-D526N become gradually shorter and less abundant (Fig. 1C, 4-6 hr). We previously reported that the RNA methyltransferase Hen1p methylates only single-stranded scnRNAs to protect them from degradation (Kurth and Mochizuki 2009). The mid-stage disappearance of scnRNAs in Slicer-dead TWI1 cells could be a result of a lack of methylation of double-stranded scnRNAs.
~24 nt RNA bound to FLAG-HA-Twi1p-D526N in the mid stages of conjugation (4-6 hr post-mix) (Fig. 1C, arrow) and accumulated (Fig. 1D, arrow). Northern blot analysis demonstrated that ~28-29 nt scnRNAs, but not the ~24 nt RNAs, hybridize to a Tlr1-1 oligo DNA probe, which is complementary to a subset of scnRNAs derived from repeated Tlr1 IES elements (Fig. S1O). Therefore, the ~24 nt RNAs probably are not degradation products of scnRNAs but are likely constitutively expressed ~23-24 nt siRNAs (Lee and Collins 2006). These ~23-24 nt siRNAs might mis-associate with Twi1p and therefore be stabilized when scnRNAs are reduced.
The localization of FLAG-HA-Twi1p-WT and FLAG-HA-Twi1p-D526N was analyzed by indirect immunofluorescence staining using an anti-HA antibody. FLAG-HA-Twi1p-WT localized to both the cytoplasm and the parental macronucleus during early stages of conjugation (Fig. 2A). During mid stages it localized almost exclusively to the parental macronucleus (Fig. 2B). In the later stages of conjugation, FLAG-HA-Twi1p-WT disappeared from the parental macronucleus and appeared in the newly developing macronucleus (Fig. 2C, D). This localization pattern was indistinguishable from that of non-tagged wild type Twi1p detected with an anti-Twi1p antibody (see Fig. 5A-C), indicating that the presence of the FLAG-HA tag did not disturb the localization of Twi1p.
In contrast, FLAG-HA-Twi1p-D526N was detected in the cytoplasm throughout conjugation and did not accumulate in the parental macronucleus (Fig. 2E-G). In the late stages of conjugation, FLAG-HA-Twi1p-D526N accumulated at the periphery of the newly developing macronuclei, but was still mostly seen in the cytoplasm (Fig.2H). These results were confirmed by analyzing nuclear and cytoplasmic fractions by western blotting (Fig. 2J). Though FLAG-HA-Twi1p-WT was mainly detected in the nuclear fraction, FLAG-HA-Twi1p-D526N was mainly detected in the cytoplasmic fraction. In contrast, the macronuclear protein Pdd1p (Coyne et al. 1999) was detected in the nuclear fraction in both FLAG-HA-TWI1-WT and FLAG-HA-TWI1-D526N strains. The other Slicer-dead Twi1p mutant (FLAG-HA-Twi1p-H745Q, Fig. S1E, F) also localized to the cytoplasm (Fig. 2I, Fig. S2). These results indicate that Slicer activity is required for the Twi1p nuclear localization.
Since comparable levels of scnRNAs accumulate in both wild type and Slicer-dead FLAG-HA-TWI1-D526N strains at 4 hr post-mixing (Fig. 1D), at which point wild type Twi1p was already localized to the parental macronucleus (Fig. 2B), the mislocalization of Slicer-dead Twi1p was not likely due to the nucleolytic shortening or reduction of scnRNAs in the mutants, but instead probably is directly caused by defective passenger-strand removal of scnRNAs.
Two possible mechanisms could explain Slicer-dependent Twi1p nuclear localization. The first suggests that Twi1p complexed with single-stranded scnRNA is anchored in the nucleus through an interaction between scnRNA and nascent macronuclear noncoding transcripts (Aronica et al. 2008). However, this anchoring cannot fully explain the nuclear localization of Twi1p because EMA1 KO strains, in which the scnRNA-noncoding RNA interaction is impaired, show normal Twi1p macronuclear localization (Aronica et al. 2008). The second and more likely possibility is that an active macronuclear import mechanism specifically recognizes the complex formed between Twi1p and single-stranded scnRNA.
None of the previously identified Twi1p-associated proteins are required for macronuclear localization of Twi1p (Aronica et al. 2008, Bednenko et al. 2009). To identify Twi1p-associated proteins involved in the macronuclear localization of Twi1p, FLAG-HA-Twi1p-containing complexes were isolated with milder lysis/washing conditions (see Materials and Methods) than those used in the previous studies. Immunoprecipitated samples from cells at the mid (5 hr post-mixing) and the late (9 hr) stages of conjugation were separated by SDS-PAGE and analyzed by silver staining (Fig. 3A). In addition to a band corresponding to FLAG-HA-Twi1p, a previously unidentified ~115 kDa protein was detected in FLAG-HA-TWI1 strains but not in non-tagged, wild-type strains. In this study, the three previously identified Twi1p-associated proteins [Ema1p (211 kDa), CnjBp (200 kDa) and Wag1p (123 kDa)] were undetectable by silver-staining although Ema1p and Wag1p were weakly detectable by western blot (Fig. S3A), most likely due to the milder lysis conditions employed. The milder lysis procedure used here solubilizes mainly cytoplasmic components including the 115kDa protein (see below), whereas all three previously identified Twi1p-associated proteins localize mainly to nuclei (Aronica et al. 2008, Bednenko et al. 2009) and require harsher lysis conditions to be observed.
We identified the 115 kDa protein by mass spectrometry (Fig. S3B) and named it Giw1p (gentleman-in-waiting). The molecular weight of Giw1p as predicted from the GIW1 mRNA sequence (GenBank XM_001029843) is 125 kDa. Giw1p shows no obvious similarity with any previously identified protein from any organism.
The interaction between Twi1p and Giw1p was confirmed by co-immunoprecipitation using GIW1-HA strains in which all of the macronuclear GIW1 loci were replaced by a GIW1-HA construct encoding C-terminal HA-tagged Giw1p (Fig. S1H, I). GIW1-HA can replace essential function (see below) of GIW1 in the production of sexual progeny (Fig. S1L), indicating that Giw1p-HA was functional and retained normal Giw1p physical interactions with other molecules. Two GIW1-HA or two non-tagged strains were crossed, Giw1p-HA-containing complexes were immunoprecipitated with an anti-HA antibody, and the precipitated proteins were analyzed by western blot using an anti-Twi1p antibody. As shown in Fig. 3B, a substantially higher amount of Twi1p was precipitated from the GIW1-HA strains than from the non-tagged strains at all developmental stages tested, confirming that Twi1p and Giw1p are found in the same complex. Silver-staining of these precipitaed proteins detected only two specific proteins with the sizes of Giw1p-HA and Twi1p (Fig. S3C), suggesting that Giw1p may complex only with Twi1p.
Since Twi1p associates with long noncoding RNAs (ncRNA) (Aronica et al. 2008), we determined whether the interaction between Twi1p and Giw1p was mediated by ncRNA. Lysates from FLAG-HA-TWI1-WT cells at 4 hr post-mixing were incubated with 20 pg/mL of RNase A to degrade ncRNAs, and the Twi1p-Giw1p interaction was analyzed by immunoprecipitation using an anti-FLAG antibody. The amount of Giw1p co-immunoprecipitated with FLAG-HA-Twi1p was comparable with (+) and without (-) RNaseA treatment (Fig. 3C, Giw1p), while ncRNA was undetectable by RT-PCR in the immunoprecipitated sample from the RNase-treated lysate (Fig. 3C, ncRNA RT+). These data suggest that the interaction between Twi1p and Giw1p is not mediated by long ncRNAs. This conclusion is further supported by the fact that Giw1p was co-immunoprecipitated with Twi1p from EMA1 KO strains (Fig. S3D), in which the Twi1p-ncRNA interaction is impaired (Aronica et al. 2008).
In the conditions described above, the amount of scnRNAs was unchanged after RNase A (20 pg/mL) treatment (+ in Fig. 3C). However, scnRNAs were eliminated when we treated the lysate with a much higher concentration (100 μg/mL) of RNase A (++ in Fig. 3C). Even in this condition, a significant, albeit reduced amount (~60%) of Giw1p was co-precipitated with FLAG-HA-Twi1p (Fig. 3C). This result suggests that Giw1p can interact with Twi1p in the absence of scnRNA in cell lysate. This conclusion is further supported by a GST-pull down assay using recombinant Twi1p expressed in E. coli and in vitro translated Giw1p. Giw1p was co-precipitated with full length Twi1p fused with GST but not with GST alone (Fig. 3D). Treatment with 100 μg/mL RNase A did not affect precipitation of Giw1p with GST-Twi1p (Fig. 3D) suggesting that contaminating RNA does not mediate interaction of these two proteins. We conclude that Giw1p and Twi1p interact directly without RNA.
Twi1p shares conserved PAZ and Piwi domains with other Argonaute proteins (Mochizuki et al. 2002). To determine the domain(s) of Twi1p that interacts with Giw1p, GST-pull down assays were performed using Giw1p and N-terminal, PAZ, Mid, Piwi or C-terminal domains of Twi1p, each fused with GST. Giw1p co-precipitated with the PAZ and the Piwi domains but not with other domains of Twi1p (Fig. 3D), indicating that Twi1p directly interacts with Giw1p through its PAZ and Piwi domains.
To determine which parts of Giw1p mediate the interaction with these domains, Giw1p was divided into six segments (Fig. 3E), all of which were examined for binding with N-terminal, PAZ, Mid or Piwi domains of Twi1p. Three of the six segments of Giw1p (N1, N3 and C2) were efficiently co-precipitated with PAZ and Piwi domains of Twi1p but less efficiently with N-terminal and Mid domains (Fig. 3E). Small amounts of the other three segments (N2, C1 and C3) were also co-precipitated with PAZ and Piwi domains, while a part of β-lactamase, which was used as a negative control, was not (Fig. 3E), suggesting that these Giw1p segments also have binding activity, albeit weak, to PAZ and Piwi domains of Twi1p. Interaction between the N3 fragment of Giw1p and PAZ/Piwi domains of Twi1p was further confirmed by a reverse GST-pull down assay using GST tagged Giw1p-N3 and His tagged PAZ/Piwi domains (Fig. S3E). His-PAZ and His-Piwi were co-precipitated with GST-Giw1p-N3 but not with GST alone. These results indicate that Giw1p has several different sites which have the ability to bind PAZ and Piwi domains of Twi1p and could bridge these domains.
Like TWI1 mRNA (Mochizuki et al., 2002), GIW1 mRNA expression occurs exclusively during early conjugation stages (2-4 hr post mixing) but was not detected in exponentially growing or starved vegetative cells (Fig. 4A).
To study the expression and localization of Giw1p, two GIW1-HA strains were crossed and Giw1p-HA was detected using an anti-HA antibody. Giw1p-HA was specifically detected during conjugation by western blotting (Fig. 4A). Indirect immunofluorescent staining showed that Giw1p-HA was localized to both the cytoplasm and nuclei throughout conjugation (Fig. 4B-F).
To elucidate the function of Giw1p, GIW1 knockout (KO) strains were constructed. All copies of the GIW1 gene in the polyploid macronucleus were replaced by genes in which the entire coding sequence had been replaced by a drug resistance marker (see Fig. S1K, L).
Two wild type or two GIW1 KO strains were mated, and the localization of Twi1p was analyzed by indirect immunofluorescence staining using an anti-Twi1p antibody. In wild type cells, Twi1p was detected mainly in parental (Fig. 5A, B) or newly developing macronuclei (Fig. 5C), whereas in GIW1 KO cells, Twi1p localized to the cytoplasm throughout conjugation (Fig. 5D-F). These data indicate that Giw1p is required for nuclear localization of Twi1p.
We also analyzed the localization of Ema1p, Pdd1p and Wag1p, which show similar localization patterns as Twi1p in wild type cells (Fig. S4). All of these proteins localize to macronuclei in GIW1 KO cells (Fig. S4), indicating that Giw1p is not a general nuclear transporter, but is dedicated to Twi1p or to a limited set of proteins.
Since Slicer-dead TWI1 and GIW1 KO strains showed a similar nuclear Twi1p localization defect, Giw1p could have a function in passenger-strand removal of the Twi1p-associated scnRNAs. To assess this possibility, Twi1p was immunoprecipitated from wild type and GIW1 KO strains using an anti-Twi1p antibody, and Twi1p-associated scnRNA was analyzed. Denaturing gel analysis indicated that similar amounts of scnRNA were associated with Twi1p in the presence and absence of Giw1p (Fig. 6A). Native gel analysis detected little double-stranded scnRNA in the absence of GIW1 (Fig. 6B). We therefore conclude that Giw1p is not required for the production, loading or passenger strand removal of scnRNAs and that Giw1p most likely acts downstream of scnRNA passenger strand removal in the pathway of Twi1p macronuclear import.
To understand the relationship between the slicing of the scnRNA passenger-strand and the action of Giw1p, two wild-type FLAG-HA-TWI1-WT strains or two Slicer-dead FLAG-HA-TWI1-D526N strains were mated, and FLAG-HA-Twi1p-WT or FLAG-HA-Twi1p-D526N were immunoprecipitated using an anti-FLAG antibody. The co-immunoprecipitation of Giw1p was analyzed by western blot using an anti-Giw1p antibody. As shown in Fig. 6C, Giw1p was co-immunoprecipitated with FLAG-HA-Twi1p-WT (WT), while no detectable Giw1p was precipitated with FLAG-HA-Twi1p-D526N (D526N). Similar results were obtained using the other Slicer-dead mutant FLAG-HA-TWI1-H745Q (Fig. S5). These results indicate that the Slicer-activity of Twi1p has essential role in the Twi1p-Giw1p interaction in vivo. Because Giw1p is required for the macronucelar localization of Twi1p (Fig. 5), the lack of interaction between the Slicer-dead Twi1p mutants and Giw1p explains why Twi1p macronucelar localization is inhibited in the Slicer-dead TWI1 strains. Since Slicer-activity of Twi1p is important for the passenger-strand removal of scnRNAs, the inability of Slicer-dead Twi1p mutants to interact with Giw1p in vivo is likely caused by the association of double-stranded scnRNAs with these mutants.
To test this hypothesis, the effect of double-stranded scnRNAs on Twi1p-Giw1p interaction was analyzed in vitro (Fig. 6D). Recombinant GST alone (lane 1) or wild type Twi1p fused to GST (GST-Twi1p) (lanes 2-4) was first incubated with (lanes 3, 4) or without (lanes 1, 2) a 28 nt guide RNA, then with (lane 4) or without (lanes 1-3) a phosphorothioate-modified non-cleavable 28 nt target (passenger) RNA. Then, GST-pull down assays were performed with the radiolabeled Giw1p. The amount of Giw1p co-precipitated with GST-Twi1p was greatly reduced when GST-Twi1p was pre-incubated with a guide and a non-cleavable target RNA (Fig. 6D, lane 4), but not with a guide RNA alone (lane 3), indicating that the presence of double-stranded scnRNAs in Twi1p inhibits the Giw1p-Twi1p interaction. These results suggest that Giw1p can sense the state of the scnRNA complexed with Twi1p and binds to both unloaded Twi1p and Twi1p that is associated with single-stranded scnRNA.
Since Giw1p can bind to Twi1p without scnRNA in a cell lysate (Fig. 3C) and in vitro (Fig. 3D), it seems reasonable to expect that unloaded Twi1p could localize in the macronucleus. However, Twi1p was localized to the cytoplasm in the absence of DCL1 (Fig. 5G), in which no detectable scnRNAs are produced (Malone et al. 2005, Mochizuki and Gorovsky 2005). One possibility is that loaded single-stranded scnRNA may be required for binding to the nuclear import machinery. Alternatively, as a significant proportion of Twi1p that enters into the macronucleus in a wild type cell is predicted to be released from scnRNA by the selective degradation of scnRNAs complementary to the macronuclear DNA (Mochizuki and Gorovsky 2004), there may be a mechanism that exports unloaded Twi1p to the cytoplasm where Twi1p could load a new scnRNA cargo, thereby preventing accumulation of unloaded Twi1p in the nucleus.
In both the Slicer-dead TWI1 and GIW1 KO strains, Twi1p is not localized to the developing macronucleus (Fig. 2G, H, Fig. 5F) where the Twi1p-scnRNA complexes are required for DNA elimination. We studied DNA elimination at four different loci by single progeny PCR (Fig. 7A) and found that their eliminations were indeed inhibited in the progeny of Slicer-dead FLAG-HA-TWI1-D526N strains (Fig. 7B).
Because most of the GIW1 KO cells are blocked at the mid-stage of conjugation (see below), it was difficult to study their DNA elimination by PCR. Instead, we used fluorescence in situ hybridization (FISH) to analyze DNA elimination of the Tlr1 and the REP IES elements, which are moderately repeated in the micronuclear genome (Wuitschick et al. 2002, Fillingham et al. 2004). Both elements remained present in the new macronucleus of most of the progeny of GIW1 KO, as well as of FLAG-HA-TWI1-D526N strains at 36 h post-mixing but were completely removed in the progeny of wild type cells (Fig. 7C). Therefore, the absence of Giw1p inhibits DNA elimination of these IES elements.
Like all other known mutants showing defective DNA elimination, Slicer-dead TWI1 and GIW1 KO strains did not produce viable mating progeny (Fig. S1M). In addition, GIW1 KO strains showed developmental arrest and ~70% of cells aborted mating at mid conjugation (Fig. S6). This phenotype was not observed in TWI1 KO strains (Mochizuki and Gorovsky 2004) and Slicer-dead TWI1 mutants (data not shown). Since scnRNAs are believed to be derived from genic as well as nongenic sequences, they potentially target many different mRNAs for degradation if they are not properly regulated. Thus, the pleiotropic defects in GIW1 KO cells could be due to the presence of mature Twi1p-scnRNA complexes in the cytoplasm. Alternatively, because Giw1p bind to the PAZ and Piwi domains of Twi1p (Fig. 3H), Giw1p might block Twi1p-associated scnRNAs from binding to mRNAs (or other RNAs) or directly inhibit Slicer activity of Twi1p.
In this study we have shown that the Tetrahymena Argonaute protein Twi1p has Slicer activity and that this activity is essential for its macronuclear localization. We have also identified a novel Twi1p-associated protein, Giw1p, which is required for macronuclear localization of Twi1p. Giw1p binds to wild-type Twi1p, but not to Slicer-dead Twi1p in vivo. These results indicate that Slicer-dependent passenger-strand removal of scnRNAs is a prerequisite for the Twi1p-Giw1p interaction, which, in turn, is essential for the macronuclear localization of Twi1p. Thus, Giw1p serves as a gatekeeper that allows only mature Twi1p-scnRNA complexes to enter macronuclei. Currently, the detailed mechanism by which Giw1p function is not clear. The simplest hypothesis is that Giw1p might be an adapter protein that connects the Twi1p-scnRNA complex to nuclear import machinery.
Since the nuclear localization of some Argonaute proteins is dependent on the presence of their small RNA cargos in nematodes (Guang et al. 2006) and in mice (Aravin et al. 2008), small RNA-dependent nuclear localization of Argonaute proteins is probably widespread among eukaryotes. This study reveals yet another layer of the regulatory mechanisms for the nuclear localization of small RNA-Argonaute complexes: the requirement for passenger-strand removal for the nuclear import of a small RNA-Argonaute complex. This mechanism might have evolved to provide Argonaute proteins enough time to release aberrant RNAs and to find correct RNAs before they are imported into the nucleus. Alternatively, proteins that block or modulate the activity of mature small RNA-Argonaute complexes during their transport might have evolved first and then later may have acquired a direct role in the nuclear import process. Since it is not yet known whether maturation of Argonaute-small RNA complexes is required for nuclear transport of Argonaute proteins in other eukaryotes, the localization of Slicer-dead Argonautes will be of interest to study in other systems.
The conformation of a bacterial Argonaute protein changes according to the state of nucleic acids with which the protein is complexed, such that the space between the PAZ and Piwi domains is wider when it is associated with both guide- and substrate-strands than when it is associated with only a guide-strand (Wang et al. 2008). Since Giw1p binds to both the PAZ and the Piwi domains of Twi1p (Fig. 3D), we propose that scnRNA passenger strand removal also alters the distance between the PAZ and Piwi domains, allowing binding of Giw1p. As Giw1p is the only currently known protein that can detect the state of small RNAs (double- or single-stranded) associated with Argonaute proteins, identification of structural/functional homologs of Giw1p in other eukaryotes could aid in understanding how conformational changes of Argonaute proteins affect their functions. Since Giw1p shows no obvious similarity with any previously identified proteins, determination of its crystal structure should prove valuable in identifying such structural/functional homologs and in elucidating mechanisms of this process.
Tetrahymena strains, culture conditions, DNA elimination assay, progeny viability assay and oligonucleotide used are described in Supplemental Data.
GST- or His-tagged Twi1p, hAgo1, Giw1p were expressed in E. coli. 35S-labeled full length and parts of Giw1p were synthesized by a in vitro translation system. See Supplemental Data for the detailed procedures.
~3 pmol of the recombinant GST or GST fusion proteins were pre-incubated with 3 pmol of 27 mer or 21 mer guide RNAs (27 mer for Twi1p and 21 mer for hAgo2) in 30 μl of 1× cleavage buffer [30 mM HEPES (pH 7.4), 40 mM KOAc, 5 mM Mg(OAc)2, 5 mM DTT] containing 1 μg BSA, 0.5 μg yeast RNA (Ambion) and 40 units RNasin (Promega) for 90 min at 25°C (for Twi1p) or 37°C (for hAgo2). 27 mer (for Twi1p) or 21 mer (for hAgo2) 32P-labeled Target RNA was added and incubated for 90 min at 25°C (for Twi1p) or 37°C (for hAgo2). RNA was extracted with phenol/chloroform followed by ethanol precipitation, separated in a 20 % denaturing polyacrylamide gel and analyzed by autoradiography.
Rabbit anti-Giw1p and anti-Wag1p antibodies were produced using synthetic peptides. See Supplemental Data for the detailed procedures.
2 × 106 cells were lysed by sonication in 1 ml lysis buffer A [20 mM Tris pH 7.5, 100mM NaCl, 2 mM MgCl2, 2 mM CaCl2, 1% Tween 20, 0.1 mM PMSF, 1× complete protease inhibitor cocktail (Roche) and 0.4 U/ml RNasin (Promega)]. For Giw1p-HA immunoprecipitation, lysis buffer B (buffer A w/o Tween 20) was used. FLAG-HA-Twi1p, Giw1p-HA or Twi1p complex was immunoprecipitated using anti-FLAG (M2, Sigma), anti-HA (HA-7, Sigma) or anti-Twi1p (Aronica et al., 2008) antibody, respectively. FLAG-HA-Twi1p complexes were eluted in 0.3 mg/ml 3× FLAG peptide (Sigma). RNA in the eluate was extracted by TRIzol. Giw1p-HA and Twi1p complexes were eluted by boiling the gels in SDS-PAGE sample buffer or by incubating gels in the TRIzol. RNAs were separated in 15% denaturing or on 18% native polyacrylamide gels and were detected either directly by GelRed (Gentauer) or by northen blot (Aronica et al. 2008) probed with 5′end radiolabeled oligo DNAs (M-28nt, Tlr1-28nt or Tlr1-1). ncRNA was analyzed by RT-PCR (Aronica et al, 2008). FLAG-HA-Twi1p, Giw1p and Twi1p were detected by western blot using anti-FLAG, anti-Giw1p and anti-Twi1p antibodies, respectively.
Cells were fixed and processed as described (Loidl and Scherthan 2004). See Supplemental Data for the detailed procedures.
A pellet of 2×106 cells was gently resuspend in the ice-cold 1 mL lysis buffer [10 mM Tris pH 7.5, 5 mM MgCl2, 10 mM KCl, 0.05 % Triton X-100, 1× complete protease inhibitor cocktail (Roche)] and immediately centrifuged at 3,000 rpm for 5 min at 4°C. The supernatant was mixed with an equal volume of 2× SDS-PAGE sample buffer. The pellet was resuspended with the lysis buffer to make final volume 1 mL and was mixed with an equal volume of 2× SDS-PAGE sample buffer.
FLAG-HA-TWI1 strains at 5 or 9 hr post-mixing were lysed with a Dounce homogenizer in lysis buffer B, proteins were immunoprecipitated with anti-HA (HA-7) agarose, and eluted with 0.2 mg/mL HA peptide (Sigma). The eluate was separated by SDS-PAGE and visualized by silver or Coomassie Blue staining. The Coomassie Blue-stained ~115 kDa band was analyzed as described (Bowman et al. 2005).
For the experiment shown in Fig. 3D and E, GST, GST-Twi1p, or GST-Giw1p-N3 (~ 1 μg) was incubated with 20 μl glutathione sepharose 4B resin (GE Healthcare) in GST pull-down buffer (GPB) (20 mM Tris-HCl (pH 7.5), 100 mM NaCl, 2 mM MgCl2, 2 mM CaCl2, 0.1% BSA) for 30 min at 4°C. For the experiment shown in Fig. 6D, GST or GST-Twi1p (~ 2 μg) in PBS with 5 mM Mg(OAc)2, 5 mM DTT and ribolock RNase inhibitor (Fermentas) were incubated with or without 4.4 nmol 28 nt guide RNA for 90 min at 26°C. Then, 17-22 nmol of 28 nt non-cleavable target RNA, 24 nt or 28 nt non-target RNA (both provided similar results) or water were added and the reaction was incubated for 90 min at 26°C. 20 μl glutathione sepharose 4B resin in GPB was added and the reactions were incubated for 30 min at 4°C. The beads were washed with GPB and incubated with 35S-labeled full-length or parts of Giw1p recombinant protein (1.2 to 2 μl reaction of in vitro translation), or with His-tagged PAZ-, mid- or Piwi-domain of Twi1p (~0.4 μg) in GPB for 90-120 min at 4°C. The beads were washed with GPB, boiled in SDS-PAGE sample buffer, and the elutions were separated by SDS-PAGE. 35S-labeled proteins were detected by phosphorimager (GE Healthcare) or by autoradiography. His-tagged proteins were detected by western blot using an anti-His antibody (Qiagen).
We thank Leemor Joshua-Tor for E. coli strain expressing GST-hAgo2, Kathy Karrer for Tlr1 clones, and Tim Clausen, Javier Martinez, Thomas Marlovits, Josef Loidl, Stefan Westermann and their laboratories for materials and technical advice. The research leading to these results received funding from the European Research Council (ERC) the Marie Curie Action “Early Stage Training” (MEST-CE-2005-019676) under the European Community's Sixth Framework Programme to HMK and LA, from the Ontario Research and Development Challenge Fund and MDS SCIEX to KWMS, from a Canadian Institutes for Health Research grant (MOP13347) to REP, from National Institutes of Health grants GM21793 and GM72752 to MAG, from ERC Starting Grant (204986) under the European Community's Seventh Framework Programme and from the Austrian Academy of Sciences to KM.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.