MicroRNAs (miRNAs) regulate plant development by post-transcriptional regulation of target genes. In Arabidopsis thaliana, DCL1 processes precursors (pri-miRNAs) to miRNA duplexes, which associate with AGO1. Additional proteins act in concert with DCL1 (e.g. HYL1 and SERRATE) or AGO1 to facilitate efficient and precise pri-miRNA processing and miRNA loading, respectively. In this study, we show that the accumulation of plant microRNAs depends on RECEPTOR FOR ACTIVATED C KINASE 1 (RACK1), a scaffold protein that is found in all higher eukaryotes. miRNA levels are reduced in rack1 mutants, and our data suggest that RACK1 affects the microRNA pathway via several distinct mechanisms involving direct interactions with known microRNA factors: RACK1 ensures the accumulation and processing of some pri-miRNAs, directly interacts with SERRATE and is part of an AGO1 complex. As a result, mutations in RACK1 lead to over-accumulation of miRNA target mRNAs, which are important for ABA responses and phyllotaxy, for example. In conclusion, our study identified complex functioning of RACK1 proteins in the Arabidopsis miRNA pathway; these proteins are important for miRNA production and therefore plant development.
MicroRNAs (miRNAs) in both animals and plants are transcribed as longer primary microRNAs (pri-miRNAs) from which RNAse III-like DICER enzymes release miRNAs of 20–22 nt length. Mature miRNAs are incorporated into an ARGONAUTE (AGO) effector protein to create an RNA-induced silencing complex (RISC) that mainly regulates target mRNAs post-transcriptionally (Voinnet, 2009). miRNA-mediated control of gene expression is essential for many aspects of plant development, including root and leaf development, hormone responses, developmental transitions and stress responses (Voinnet, 2009).
In the model plant Arabidopsis thaliana, one of the four DICER-LIKE (DCL) proteins, DCL1, is mainly involved in the production of miRNAs (Park et al., 2002; Fahlgren et al., 2009; Laubinger et al., 2010). The proteins DAWDLE (DDL), TOUGH (TGH), HYPONASTIC LEAVES 1 (HYL1) and SERRATE (SE) interact with DCL1 and are responsible for precise and efficient miRNA production (Vazquez et al., 2004; Hiraguri et al., 2005; Kurihara et al., 2006; Lobbes et al., 2006; Yang et al., 2006; Dong et al., 2008; Yu et al., 2008; Ren et al., 2012). During miRNA production, long primary miRNA transcripts (pri-miRNAs) are cut to form precursor miRNAs (pre-miRNAs), from which miRNA duplexes are released. miRNAs are then methylated by HUA ENHANCER 1 (HEN1) and associate mainly with AGO1, one of the ten AGO proteins in Arabidopsis (Yu et al., 2005; Mi et al., 2008; Montgomery et al., 2008). HEAT SHOCK PROTEIN 90 (HSP90) and the cyclophilin 40 protein SQUINT (SQN) physically interact with AGO1 and ensure efficient miRNA loading (Smith et al., 2009; Iki et al., 2010, 2012; Earley and Poethig, 2011). AGO1 removes the passenger strand (or miRNA*) of the miRNA duplex, allowing the miRNA guide strand to recognize complementary sequence regions within target RNAs (Mallory et al., 2008). Binding of AGO1 to target sequences results in mRNA cleavage within the miRNA binding site or translational inhibition occurring on the endoplasmic reticulum (Voinnet, 2009; Li et al., 2013). Because miRNAs play such important roles in plant development, mutants impaired in the production of miRNAs or in their function exhibit a wide range of developmental defects, and null alleles of DCL1 or SE are embryonically lethal (Lobbes et al., 2006; Schauer et al. 2002). Over recent years, additional small RNA (sRNA) factors have been identified that are not solely involved in sRNA metabolism, but possess pleiotropic functions in diverse biological pathways. Among them are RNA polymerase II, the MEDIATOR complex, CPL phosphatases, the importin β protein EMA1/SAD2, KATANIN and proteins involved in the isoprenoid biosynthesis pathway (Brodersen et al., 2008, 2012; Kim et al., 2011; Wang et al., 2011b; Hajheidari et al., 2012; Manavella et al., 2012). These examples indicate that some miRNA factors have already been reported to operate in diverse biological processes.
In order to identify additional components of the plant miRNA pathway, we performed a yeast two-hybrid screen using the pri-miRNA processing factor SE as bait, and identified RECEPTOR OF ACTIVATED C KINASE 1 (RACK1) as a potential interactor. RACK1 is present in all eukaryotic organisms studied, and possesses seven WD40-β-propellers, which mediate simultaneous interactions with multiple proteins (Nilsson et al., 2004; Ullah et al., 2008; Adams et al., 2011). RACK1 itself has no enzymatic activity, and instead acts as a bridge for interactions between proteins or competes with other proteins for binding pockets, thereby inhibiting specific interactions. RACK1 is often identified in proteomic approaches, suggesting that RACK1 is part of many complexes, in which it controls complex formation and stability, or creates a docking site for regulators (Gibson, 2012). Several types of direct RACK1 interaction partners have been described (Nilsson et al., 2004). RACK1 is a core component of the eukaryotic 40S ribosomal subunit, and is thought to directly regulate translation in response to various stimuli (Nilsson et al., 2004; Adams et al., 2011). In the nucleus, RACK1 acts as adaptor to connect kinases with their substrates, and modulates transcription (He et al., 2002; Nery et al., 2004; Wang et al., 2011a; Neasta et al., 2012). Animal RACK1 has been shown to act in the miRNA pathway. The Caenorhabditis elegans Argonaute protein that is involved in miRNA action, ALG-1, directly binds to ribosome-associated RACK-1, implying that C. elegans RACK-1 may be an anchor point for ALG-1 on mRNAs that are being translated (Jannot et al., 2011). In another study using a liver tumor cell line, RACK1 was shown to affect loading of miRNAs into the effector complex and to regulate the localization of KH-TYPE SPLICING REGULATORY PROTEIN (KSRP), which promotes maturation of a small subset of pre-miRNAs (Otsuka et al., 2011). These contradictory results are further complicated by the observation that RACK1 knockdown in C. elegans, but not in human cells, results in higher miRNA levels (Jannot et al., 2011; Otsuka et al., 2011). More specific functions of RACK1 in the miRNA pathway may be obscured by its pleiotropic actions in many diverse cellular processes.
In contrast to animals and yeast, very little is known about RACK1 in plants including its function in the plant miRNA pathway. The Arabidopsis genome encodes three RACK1 genes: RACK1A, RACK1B and RACK1C (Guo and Chen, 2008). Previous analyses of single and multiple mutants revealed that RACK1 genes act redundantly to control hormone signaling, leaf development and root growth (Chen et al., 2006; Guo and Chen, 2008; Guo et al., 2009). RACK1A plays a predominant role: it is expressed most highly, and mutations in RACK1A cause slight phenotypic abnormalities, while rack1b and rack1c single mutants are indistinguishable from wild-type siblings (Chen et al., 2006; Guo and Chen, 2008; Guo et al., 2009). rack1 triple mutants exhibit strong pleiotropic defects and eventually die before producing seeds (Guo and Chen, 2008).
Here, we show that Arabidopsis RACK1, unlike its homologs in animals, is important for miRNA accumulation. Our results show that Arabidopsis RACK1 controls correct pri-miRNA accumulation and interacts with known miRNA processing and effector complex components, suggesting that RACK1 fulfils more than one function in the plant miRNA pathway.
RACK1, an SE-interacting protein, is important for miRNA accumulation
We identified RACK1B as a potential SE interactor in a yeast two-hybrid screen. In a yeast interaction assay, all three Arabidopsis RACK1 proteins were capable of interacting with SE (Figure 1a). We subsequently determined whether RACK1 is associated with SE in planta. First, we performed co-immunoprecipitation experiments focusing on the major RACK1 gene in Arabidopsis, RACK1A. A RACK1A–GFP fusion protein expressed under the control of endogenous RACK1A regulatory sequences (RACK1A:RACK1A-GFP) rescues the rack1a abscisic acid (ABA)-hypersensitive phenotype (Guo et al., 2009). Pull-down experiments using a GFP affinity matrix showed that SE co-purified with RACK1–GFP (Figure 1b). In contrast, SE was barely detectable in pellet fractions of immunoprecipitation experiments using wild-type extracts (Figure 1b). We further confirmed the SE–RACK1 interaction by bimolecular fluorescence complementation (BiFC). SE and RACK1 interacted in the nucleus, and were often concentrated in distinct foci (Figure 1c). This is in agreement with the fact that SE mainly localizes to nuclear D-bodies (Fang and Spector, 2007), and that Arabidopsis RACK1 was present in the cytosol and nucleus in planta (Figure 1d). Some RACK1 targets have been reported to be stabilized or destabilized upon binding (Liu et al., 2007; Zhang et al., 2012), but we found that the amounts of SE remained unchanged in rack1 mutants (Figure 1e). These results suggest that RACK1 does not influence SE protein levels.
Because RACK1 interacted with the miRNA factor SE, we tested whether miRNA levels were affected in rack1 mutants. To do this, we quantified miR156, miR159, miR164 and miR166 by small RNA blot analyses and quantitative RT-PCR in rack1 and the known miRNA-related mutants se and ago1 (Figure 2a–e). se mutants are impaired in very early steps of miRNA biogenesis, and hence the levels of mature miRNAs were drastically decreased (Figure 2d). In ago1 mutants, miRNAs were presumably not efficiently loaded and were therefore less stable (Figure 2d). We observed a drastic decrease in the miRNA levels in rack1 triple mutants, in which miRNAs were reduced to 10–40% of the wild-type levels (Figure 2a,b). rack1a mutants also contained less miRNAs than the wild-type (Figure 2c,d). Analysis of rack1 single and double mutants revealed that a functional RACK1A gene resulted in wild-type miRNA levels under the conditions tested (Figure 2d), suggesting that RACK1A is the most important player controlling miRNA abundance.
We also tested whether the accumulation of miRNA* species was affected in rack1 mutants. We used ago1 mutants as genetic controls for miRNA* quantification by real-time PCR. miRNA* species accumulate in ago1 mutants because AGO1 activity is responsible for removal of the miRNA* strand after loading of the miRNA/miRNA* duplex (Eamens et al., 2009). As expected, the levels of some miRNA* species were increased in ago1 mutants, but their abundance was low in se mutants (Figure 2e). We found that rack1a mutants possessed low miRNA* levels, as in se-1 mutants. This effect was exaggerated in the rack1a rack1b double mutant (Figure 2e).
Additional expression studies demonstrated that RACK1 has no specificity for miRNAs that associate with AGO1, because the levels of miR390 and miR852, which mainly associate with AGO7 and AGO2, respectively, were strongly reduced in rack1 triple mutants (Figure 2f) (Mi et al., 2008; Montgomery et al., 2008). rack1 mutants also accumulated low levels of miR822, the processing of which is accomplished by DCL4, suggesting that the function of RACK1 is not limited to DCL1-dependent miRNAs (Rajagopalan et al., 2006; Figure 2f). Probably due to impaired miRNA expression, we found that the levels of trans-acting small-interfering RNAs (tasiRNAs) were also reduced in rack1 triple mutants (Figure 2f). Collectively, our findings demonstrate that the SE-interacting protein RACK1 is important for miRNA accumulation.
RACK1 affects the accumulation and processing precision of pri-miRNAs
To obtain further insights into the function of RACK1 in the plant miRNA pathway, we sequenced small RNA populations from two biological replicates of wild-type and rack1 triple mutant seedlings using Illumina technology. After removal of reads smaller than 17 nt or larger than 27 nt and reads mapping to tRNAs and rRNAs, between 5 and 7 million reads remained for each sample. First, we performed an expression analysis to estimate the global effects of RACK1 on the miRNA transcriptome. A total of 209 known miRNAs were included in our analysis, of which 77 miRNAs were represented by at least five sequencing reads in all samples, which allows robust calculation of differential expression. Of these 77 miRNAs, 61 miRNAs were significantly less abundant in rack1 triple mutants compared to wild-type (P value <0.05, false discovery rate <0.02, Figure 3a and Figure S1). Interestingly, the levels of two miRNAs, miR827 and miR866-5p, were higher in rack1 triple mutants, which may be explained by RACK1's pleiotropic functions. In general, our results suggest that RACK1 function is important for the accumulation of a large number of Arabidopsis miRNAs.
The low levels of miRNAs in rack1 may be explained by low transcriptional levels of MIRNA genes (as in mediator mutants), less stable pri-miRNAs (as observed in ddl mutants) or less efficient or mis-processing of pri-miRNAs (as observed in se or hyl1 mutants) (Dong et al., 2008; Yu et al., 2008; Kim et al., 2011; Manavella et al., 2012). In order to determine whether rack1 mutants exhibit processing defects and accumulate non-canonical, unusually sized miRNAs, we mapped all sequenced 17–27 nt sRNAs to the annotated mature miRNAs. As exemplified by miR157a–c and miR319a/b species, we found that, in rack1 mutants, most MIRNA loci released small RNAs with a similar size distribution to that in the wild-type (Figure 3b and Figure S2). However, a small subset of MIRNA loci produced an array of aberrant non-canonical miRNAs, which were absent or less abundant in the wild-type. In the case of miR159b and miR167a/b, the production of 19 and 20 nt miRNA species was moderately increased in rack1 mutants (Figure 3b). Similarly, the miR167d and miR393a species were more often 21 nt long in rack1 mutants, compared to the predominant 22 nt long miRNA found in the wild-type (Figure 3b). These observations suggest a role for RACK1 in precise processing of some pri-miRNAs, similar to the known factors SE and HYL1. We also analyzed whether pri-miRNAs were inaccurately diced outside the miRNA duplex region in the rack1 mutant background. To test this, we mapped all sequenced RNAs to the entire precursors of miR157a, miR159b and miR319a, and found that rack1 triple mutants did not produce any major products resulting from mis-processing outside the miRNA/miRNA* region (Figure 3c).
Next, we determined whether RACK1 influences pri-miRNA contents by analyzing the steady-state levels of various pri-miRNAs by quantitative PCR (Figure 4a–c). As in other pri-miRNA processing mutants such as se-1, we found that some pri-miRNAs (such as pri-miR393a) accumulated to higher levels in rack1 triple mutants (Figure 4b,c). In contrast, rack1 triple mutants also exhibited a reduction of several pri-miRNAs, suggesting that RACK1 affects MIRNA gene transcription and/or pri-miRNA stability (Figure 4b). Some pri-miRNAs were unaffected in the rack1 triple mutants (Figure 4b), in agreement with the observation that RACK1 is not a general activator of gene expression (Guo et al., 2011).
In order to test whether RACK1 affects the processing of pre-miRNAs to mature miRNAs, we performed expression analysis using oligonucleotides specific for the pre-miRNA region. Priming the reverse transcription reaction using oligo(dT) oligonucleotides allows detection of pri-miRNAs only, while priming with random hexamers delivers information about pri-miRNAs and pre-miRNAs (Figure 4c). We did not observe dramatic differences between oligo(dT)-primed and randomly primed cDNA samples (Figure 4c), suggesting that rack1 mutants are not impaired in pre-miRNA processing.
In summary, our experiments show that RACK1 has pleiotropic functions in the miRNA pathway. RACK1 affects the processing and transcription/stability of certain pri-miRNAs, and plays a minor role in the precision of pri-miRNA processing.
Arabidopsis RACK1 is associated with AGO1-containing complexes
Our data provide evidence that Arabidopsis RACK1 is important for the production of the vast majority of miRNAs. In contrast to that, rack-1-deficient C. elegans strains contain higher amounts of miRNAs, and RACK-1 function is limited to bridging the interaction between the ribosome and the AGO miRNA effector protein (Jannot et al., 2011). Although our results suggest that Arabidopsis RACK1 functions in pri-miRNA transcription/stabilization and processing, the results of studies on human and worm RACK-1 prompted us to test whether Arabidopsis RACK1 is associated with AGO1 (Jannot et al., 2011; Otsuka et al., 2011). Therefore, we performed immunoprecipitation experiments using plants expressing RACK1A–GFP. Immunoblotting experiments using an AGO1-specific antibody revealed that AGO1 co-precipitated with RACK1–GFP (Figure 5a). The pellet fractions for GFP pull-down experiments using wild-type plants did not contain detectable amounts of AGO1 (Figure 5a). Unlike in C. elegans, the interaction between RACK1 and AGO1 is likely to be indirect, as we did not detect a direct interaction between AGO1 and RACK1 in BiFC experiments (Figure 1c).
If AGO1 and RACK1 exist in a complex, they should co-localize in cells. In order to study the subcellular location of RACK1 and AGO1, we simultaneously expressed GFP–AGO1 and mRFP-RACK1 proteins in Nicotiana benthamiana cells. RACK1 and AGO1 co-localized in both the nucleus and the cytosol, further supporting the conclusion that RACK1 is part of an AGO1 complex (Figure 5b). In order to determine whether RACK1 and AGO1 exist in complexes of the same size, we performed gel filtration experiments followed by immunoblot analyses using RACK1- and AGO1-specific antibodies. Consistent with previous reports, we detected AGO1 complexes of 200–400 kDa (Figure 5c) (Azevedo et al., 2010; Csorba et al., 2010). RACK1 was highly abundant in the same fractions, in agreement with the conclusion that RACK1 and AGO1 are part of a common complex (Figure 5c).
Because AGO1 is associated with polysomes, and RACK1 is part of the small ribosomal subunit, the interaction between AGO1 and RACK1 may be limited to ribosome-associated RACK1 and AGO1. To test whether AGO1 and RACK1 were also found in complexes other than the ribosome, we tested in which kind of complexes S14, a small ribosomal subunit, was detectable in our gel filtration experiments. The elution profile of AGO1 and RACK1 was largely distinct from that of S14-containing complexes (Figure 5c), suggesting that AGO1 and RACK1 are also present in complexes other than ribosomal complexes.
The exact function of RACK1 within AGO1 complexes remains to be elucidated, but it is important to note that RACK1 does not affect the overall integrity of AGO1-containing complexes. We tested this by performing gel filtration experiments in the rack1a rack1b double mutant, and obtained a very similar elution profile for AGO1 in wild-type and the rack1 double mutant (Figure 5c). These results suggest that AGO1 complexes are assembled and remain stable in the absence of wild-type RACK1 levels.
We also determined whether RACK1 affects the cleavage accuracy of AGO1 complexes. We found that most miRNA targets were cleaved at the expected positions, with the exception of the miR160 target ARF17 (Figure 5d). In rack1 triple mutants, the ARF17 mRNA was mainly cleaved two nucleotides downstream of the main cleavage site found in wild-type. The impaired slicing accuracy is not due to mis-processing of miR160, because another miR160-targeted mRNA, ARF16, is cleaved at the correct position in rack1 mutants (Figure 5d). Taken together, these findings suggest that, in general, RACK1 does not affect the accuracy of miRNA-guided target cleavage. However, the slicing of some targets, such as ARF17, may be either directly or indirectly influenced by RACK1.
RACK1 regulates miRNA-regulated gene expression
Mutants with defects in miRNA accumulation show deregulation of miRNA-mediated gene expression. Because rack1 mutants are impaired in miRNA accumulation, we analyzed the steady-state levels of various known target transcripts. The levels of several miRNA-targeted mRNAs rose 1.5–13-fold in the rack1 triple mutant, suggesting that RACK1 is important for the regulation of miRNA target genes (Figure 6a). Most tested target mRNAs were not affected in rack1 single mutants (exemplified by expression analysis of the miR164 target CUC1, Figure 6a), which may be explained by the genetic redundancy among the RACK1 genes.
We also investigated the targets of miR398, which is known to act via transcript cleavage and translational inhibition (Dugas and Bartel, 2008; Bouche, 2010). CSD1, CSD2 and CCS mRNA levels were increased even in rack1a single mutants, a phenotype that was completely reversed by introduction of a wild-type RACK1A copy (Figure 6b,c). Furthermore, the CSD1, CSD2 and CCS proteins were much more abundant in the rack1a rack1b double mutant, further indicating that RACK1 controls miRNA-mediated gene regulation (Figure 6b,c). rack1 triple mutants also accumulated more AGO1, the mRNA of which is targeted by miR168 (Figure 6d) (Vaucheret et al., 2004).
Because miRNA-mediated gene expression is altered in rack1 mutants, we determined whether rack1 mutants share some common phenotypes with other miRNA-related mutants. Despite RACK1 having pleiotropic functions, rack1a single mutants exhibited an ABA hypersensitivity similar to that of ago1-27 mutants (Figure 6e) (Guo et al. 2011). With low penetrance, rack1a rack1b double mutants also showed phyllotaxy defects, a phenotype that is frequently observed in other miRNA-related mutants such as se-1 (Figure 6f) (Prigge and Wagner, 2001). Taken together, our results show that rack1 mutants accumulate miRNA-targeted mRNAs and exhibit phenotypic similarities to other miRNA mutants, suggesting that RACK1 is an important regulator of miRNA-mediated control of Arabidopsis development.
Over a dozen proteins are known to be part of the miRNA pathway in plants, and specific functions such as RNA binding is assigned to most of them. Here, we have shown that a scaffold protein, RACK1, participates in the miRNA pathway.
Diverse functions of RACK1 scaffold proteins in various steps of the plant miRNA pathway
Although RACK1 plays essential roles in many cellular processes, which may obscure some of its specific functions during miRNA maturation, we were able to show that RACK1 function is essential for quantitative and qualitative accumulation of mature miRNAs. Some pri-miRNAs are less abundant in rack1 mutants, suggesting that RACK1 controls MIRNA transcription and/or pri-miRNA stability. Mammalian RACK1 directly regulates transcription of the brain-derived neurotrophic factor (BDNF) gene by physical association with a distinct promoter region and induction of acetylation of histone H4 (He etal., 2010). RACK1 also indirectly influences transcription by interfering with the DNA binding capacity or reducing the stability of certain transcription factors (Okano et al., 2006; Liu et al., 2007; Zhang et al., 2012). Whether Arabidopsis RACK1 directly associates with certain MIRNA genes or whether RACK1 regulates the activity of transcription factors that bind to the promoters of MIRNA genes is an interesting subject for future research.
Our results showed that RACK1 not only affects the transcription/stabilization of pri-miRNAs, but also participates in later steps within the miRNA pathway. RACK1 interacts with SE, a factor that controls the efficiency and precision of pri-miRNA processing. This is in agreement with the observation that, similar to se mutants, some pri-miRNAs accumulate in rack1 mutants. A small subset of pri-miRNAs were not accurately processed in the absence of RACK1, such as pri-miR167b, which is accurately diced only in the presence of SE (Dong et al., 2008). Together, these results imply that the RACK1–SE interaction is important for processing of a specific subset of pri-miRNAs. Several reports have suggested that some animal miRNA precursors require a specialized protein partner for optimal miRNA production (Viswanathan et al., 2008; Paroo et al., 2009; Trabucchi et al., 2009), and this is also likely to be the case in plants (Reyes and Chua, 2007; Laubinger et al., 2010; Jung et al., 2012). The function of RACK1 may be to build a docking site for such factors, and hence pri-miRNAs and mis-processing products accumulate in the absence of RACK1 function. Because SE and its homologs in animals are not only involved in miRNA metabolism (Laubinger et al., 2008; Andreu-Agullo et al., 2012; Gruber et al., 2012), RACK1 may also control other RNA processing events in concert with SE.
In addition to its association with SE, we found that RACK1 is associated with an AGO1 complex. Defects in AGO1-related processes in the rack1 mutant background may contribute to the strong reduction of miRNA levels, as observed in other AGO1-related mutants (Smith et al., 2009). While the exact molecular function of RACK1 within ARGONAUTE complexes in plants remains unknown, it may be hypothesized that either ribosomal or non-ribosomal RACK1 is important for mediating protein–protein interactions. For instances, Arabidopsis AGO1 associates with SQN and HSP90 only very transiently, and Iki et al. (2012) were only able to detect the interactions when they performed their experiments in the presence of a non-hydrolyzable ATP analog. A possible RACK1 function may be to bring components such as HSP90, SQN and an AGO1 complex together and transiently stabilize their interactions. Detailed studies on this possible function will require a sophisticated in vitro system such as that described by Iki et al. (2012), because higher levels of AGO1 in rack1 mutants may have compensatory effects (Figure 6d).
In conclusion, we found that RACK1 function is important for the accumulation of many miRNAs, and that the reason for this is likely to be a combination of the role of RACK1 in MIRNA gene transcription/stabilization, pri-miRNA processing and as part of an AGO1 complex.
Differences between animal and plant RACK1 function in the miRNA pathway
Human RACK1 interacts with KSRP, a splicing factor that binds to terminal loops of pri- and pre-miRNAs featuring GGG triplets, but knock-down of human RACK1 does not affect miRNA levels (Trabucchi et al., 2009; Otsuka et al., 2011). In contrast, C. elegans RACK-1 directly interacts with the Argonaute protein ALG-1, but, interestingly, the steady-state levels of two tested miRNAs (lin-4 and let-7) were increased in rack-1 RNAi lines (Jannot et al., 2011). This is in striking contrast to the results we obtained for Arabidopsis RACK1, as the levels of most miRNAs were reduced in Arabidopsis rack1 mutants. Therefore, the function of Arabidopsis RACK1 appears to be much more general compared to its animal counterparts.
Down-regulation of RACK-1 by RNAi in C. elegans led to higher protein production from miRNA-targeted mRNAs, suggesting that animal RACK1 may be involved in miRNA-mediated translational repression (Jannot et al., 2011). The authors hypothesized that ribosomal RACK1 may guide the C. elegans ALG-1 protein to the mRNA, and, in agreement with this, ALG-1 recruitment to polysomes is reduced in rack-1 RNAi lines (Jannot et al., 2011). To address the question of whether Arabidopsis RACK1 is also involved in miRNA-mediated translational repression, we investigated the targets of miR398, which is known act via this mechanism (Brodersen et al., 2008; Dugas and Bartel, 2008; Bouche, 2010). Although we found that miR398-targeted mRNAs are not more abundant in rack1a rack1b double mutants compared with rack1a mutants, we observed a drastic increase in the levels of the corresponding proteins (CSD1, CSD2 and CCS). However, these differences were accompanied by a strong reduction in miR398 levels in the rack1a rack1b double mutant, which may account for the increases in CSD1, CSD2 and CCS levels (compare Figure 2c with Figure 6b,c). As RACK1 plays an important role in the plant miRNA pathway upstream of AGO1, it will be challenging to decipher the role of plant RACK1 in miRNA-mediated translational repression in plants.
As the exact functions of RACK1 in the animal and plant miRNA pathways are difficult to study, because RACK1 acts pleiotropically and is involved in so many biological processes, we cannot entirely disregard the possibility that RACK1 controls translation of a specific miRNA factor, for example. However, in strong support of our hypothesis that RACK1 directly participates in the miRNA pathway, we found that RACK1 interacts with SE and is present within AGO1 complexes, making it rather unlikely that the miRNA defects observed in rack1 mutants are indirect consequences. Most likely, RACK1 acts as an adaptor that enhances protein–protein interactions. We did not observe interactions between RACK1 and other known miRNA factors in yeast two-hybrid experiments (Corinna Speth and Sascha Laubinger, unpublished results). Although this negative result in yeast assays does not formally rule out such interactions, these observations imply that RACK1 could bridge the interaction between known miRNA components (SE or AGO1) and yet unknown miRNA players. The fact that rack1 triple mutants are viable and still produced detectable amounts of miRNAs (Figure 2a) suggests that protein–protein interactions within the miRNA pathway also occur in the absence of RACK1, but presumably with much lower efficiency. Alternatively, additional scaffold proteins that are part of the plant miRNA pathway may compensate for the lack of RACK1 function to a certain extent.
Plant material and growth conditions
The mutants and transgenic lines used in this study, i.e. the se-1, ago1-27, rack1 mutants and RACK1A:RACK1A-GFP rack1a lines, have been described previously (Morel et al., 2002; Grigg et al., 2005; Chen et al., 2006; Guo and Chen, 2008; Guo et al., 2009). Unless otherwise stated, the following rack1 alleles were used for all experiments: rack1a-1, rackb-2, rack1c-1, rack1a-1 rack1b-2, rack1a-1 rack1c-1, rack1a-1/+ rack1b-2 rack1c-1, 35S:RACK1A rack1a-1 and RACK1A:RACK1A-GFP rack1a-1. Col-0 served as a wild-type control in all experiments.
Plants were grown on half-strength MS plates for 10 days or on soil for 21 days under long-day conditions (16 h light, 22°C). For all analyses involving rack1a,b,c triple mutants, plants were grown for 10 days on half-strength MS plates supplemented with 2% w/v sucrose.
cDNAs of all genes were PCR-amplified from reverse-transcribed RNAs using Phusion polymerase (Thermo Fisher Scientific, http://www.thermoscientific.com), and cloned into the GATEWAY®-compatible entry vectors pENTR1a® or pCR®8/GW/TOPO® (Life Technologies, http://www.lifetechnologies.com). LR Clonase™ II (Life Technologies) was used to transfer cDNAs into the destination vectors pGBKT7-DEST, pGADT7-DEST, pSPYNE-DEST, pSPYCE-DEST, pGWB652 and pGWB654 to generate in-frame fusions of AGO1, SE and RACK1A with the GAL4 DNA-binding domain (BD), the GAL4 activation domain (AD), YFPn (N-terminal part of YFP), YFPc (C-terminal part of YFP), mRFP or GFP (Walter et al., 2004; Horak et al., 2008; Nakamura et al., 2010).
Yeast two-hybrid assay
The yeast strain AH109 (James et al., 1996) was transformed using standard LiAc-based transformation according to the manufacturer's protocol (Clontech, www.clontech.com). After 3 days, 5–10 colonies from each transformation were resuspended in 10% w/v glycerol, and adjusted to an OD600 of 1. We prepared a serial 1:10 dilution of the yeast cells, spotted them onto control (–WL) or selective (–WLH) media lacking the corresponding amino acids mentioned in brackets, and grew the colonies for 4 days at 28°C.
Bimolecular fluorescence complementation (BiFC), co-localization and microscopy
AGO1, SE and RACK1A fusion proteins were transiently co-expressed in 2–3-week-old Nicotiana benthamiana leaves using standard methods (de Felippes and Weigel, 2010). After 36 h, leaf discs were analyzed by confocal microscopy using a Leica TCS SP2 confocal microscope (http://www.leica-microsystems.com). The expression of the transformed fusion proteins was analyzed by immunodetection using GFP-specific antibodies (Abcam, http://www.abcam.com/). For analysis of the subcellular localization of RACK1, roots of RACK1A:RACK1A-GFP rack1a transgenic lines were analyzed by confocal microscopy as described above.
RNA isolation and analyses
Total RNA was extracted from plants using TRIZOL® (Life Technologies). For small RNA blot analysis, 10 μg total RNA was resolved by urea PAGE, blotted onto nylon membranes (Hybond N+, Amersham, http://www.gelifesciences.com), and hybridized with 5'-radiolabeled oligonucleotides (Table S1).
For quantitative RT-PCR analyses, 200 ng to 2 μg of total RNA was treated with DNase and reverse-transcribed using a RevertAid™ first-strand cDNA synthesis kit (Thermo Fisher Scientific, http://www.thermoscientific.com). Oligo(dT) primers and specific stem-loop primers were added for reverse transcription of mRNAs and miRNAs, respectively (Table S1) (Varkonyi-Gasic et al., 2007). Quantitative PCR was performed in reactions containing SYBR Green (Thermo Fisher Scientific) on a CFX384 system (Bio-Rad, http://www.bio-rad.com/). All measurements were repeated twice with at least two biological replicates and in the presence of a standard curve of amplification. TUBULIN served as a normalization control for all experiments. All oligonucleotides are listed in Table S1.
RACE experiments were performed using Firstchoice® RLM-RACE (Life Technologies). Briefly, 1 μg total RNA or 75 ng polyA RNA was ligated to the 5′-RNA adaptor, and, after reverse transcription, PCR reactions using adaptor- and gene-specific oligonucleotides (Table S1) were performed using Phusion® polymerase or DreamTaq (Thermo Fisher Scientific). The resulting PCR products were cloned into a TOPO-TA cloning vector (Life Technologies), and 7–32 individual clones were sequenced.
Small RNA libraries were generated from total RNA of wild-type and rack1 triple mutants. Construction of libraries and sequencing using the HiSeq 2000 Illumina system (http://www.illumina.com) was performed by GATC Biotech (http://www.gatc-biotech.com/). After adaptor trimming, reads between 17 and 27 bp length in each sRNA Illumina library were mapped without mismatches against the TAIR10 genome sequence (http://www.arabidopsis.org/) using GenomeMapper (Schneeberger et al., 2009). Mapped reads were annotated according to the TAIR10 genome annotation, and reads mapping to tRNA and rRNA loci were removed. The remaining reads were mapped with no mismatches against known mature miRNA sequences (http://mpss.udel.edu/common/web/starExamples.php?SITE=at_pare). miRNA expression counts were normalized to reads per million (RPM) with the total number of reads mapping to the genome. If a read mapped to more than one miRNA from the same miRNA family, the count was divided by the number of different miRNA. All analyses were performed using custom Perl scripts, which are available upon request. Differential expression analysis was performed using the EdgeR package (Robinson et al., 2010). Sequencing data were deposited at the Gene Expression Omnibus under accession number GSE40579.
Protein co-immunoprecipitation, gel filtration and immunoblot analyses
For protein extraction, plant material was ground in liquid nitrogen, resuspended in protein extraction buffer [50 mm Tris pH 7.5, 150 mm NaCl, 10% v/v glycerol, 1 mm dithiothreitol and Complete protease inhibitor (Roche, http://www.roche.com)], and lysates were cleared by centrifugation at 16 000 g, 4°C, for 30 min. For co-immunoprecipitation, plant material was ground in liquid nitrogen, resuspended in binding buffer [50 mm Tris pH 7.5, 100 mm NaCl, 10% v/v glycerol, 100 nm MG132 (Life Sensors, http://www.lifesensors.com) and Complete protease inhibitor (Roche)] and cleared by centrifugation at 16 000 g, 4°C, 30 min, and the resulting lysates were further cleared by filtration (45 μm sterile filter). The protein concentration was determined by the Bradford assay (Bio-Rad). Each co-immunoprecipitation was performed using 5 mg of total protein according to published protocols with minor modifications (Isono and Schwechheimer, 2010). Briefly, lysates was pre-cleared using 50 μl Agarose A beads (Roche) for 20 min at 4°C on a rotator. After removal of the beads, the lysates were incubated with 30 μl GFP–TRAP (Chromotek,http://www.chromotek.com) for 3 h under gentle rotation. The beads were washed three times with 1 ml of cold washing buffer (binding buffer + 0.05% v/v Triton X-100), each step 10 min at 4°C, and were resuspended in 25 μl of 2× Laemmli buffer (4% w/v SDS, 100 mM Tris pH 7.5, 20% v/v glycerol, 200 mM DTT, 0.4 mg/ml bromophenol blue).
For gel filtration experiments, plant material was resuspended in 50 mm Tris/HCl pH 7.5, 200 mm NaCl, 5% v/v glycerol), and cleared by centrifugation (16 000 g, 4°C, 30 min). Total protein (500 μg) was loaded onto a Superdex 200 column (GE Healthcare, http://www.gelifesciences.com), and 500 μl fractions were collected. Protein samples were concentrated using 10 μl StrataClean resin (Stratagene, http://www.genomics.agilent.com), and subjected to immunoblot analysis.
All protein samples were resolved by SDS–PAGE and transferred onto nitrocellulose or poly(vinylidene difluoride) membranes (GE Healthcare). Standard immunodetection was performed using protein-specific antibodies against AGO1, CCS, CSD1/2, S14, SE or RACK1 (all Agrisera, http://www.agrisera.com), GFP (Abcam) or tubulin (Sigma-Aldrich, http://www.sigmaaldrich.com), followed by visualization using enhanced chemiluminescence and a chemiluminescence imaging system (Chemi-Smart 5000, Peqlab, http://www.peqlab.de).
We thank Jin-Gui Chen (Biosciences Division, Oak Ridge National Laboratory, Oak Ridge, USA), Klaus Harter, Jianjun Guo, Tsuyoshi Nakagawa (Department of Molecular and Functional Genomics, Center for Integrated Research in Science, Shimane University, Matsue, Japan), Herve Vaucheret (Institut Jean-Pierre Bourgin, INRA Centre de Versailles-Grignon, Versailles Cedex, France) and the Nottingham Arabidopsis Stock Centre for sharing seeds and DNA constructs, Gert Huber and his team for excellent care of our plants, Anja Hoffmann for excellent technical assistance, and Christoph Schall (Interfaculty Institute for Biochemistry, University of Tübingen, Tübingen 72076, Germany) for assistance with the gel filtration analysis. We are grateful to Rebecca Schwab and Detlef Weigel (MPI for Developmental Biology, Tübingen, Germany) for helpful comments and critical reading of the manuscript, and to Olivier Voinnet (Swiss Federal Institute of Technology (ETH), Zurich, Switzerland) for sharing unpublished results. The work was supported by the Deutsche Forschungsgemeinschaft (to S.L.), the Max Planck Society (to K.S.), the ‘Research Seed Capital Program’ initiated by the University of Tübingen and the State of Baden-Württemberg (to S.L.), and the Max Planck Society Chemical Genomics Centre (CGC) through its supporting companies AstraZeneca, Bayer CropScience, Bayer Healthcare, Boehringer-Ingelheim and Merck (to S.L).