Present address: Department of Genetics and Molecular Biology, University of Rome ‘La Sapienza’, Rome, Italy.
These two authors share first authorship.
Short interfering (si) and micro (mi) RNAs influence gene expression at post-transcriptional level. In plants, different classes of DICER-LIKE (DCL) enzymes are responsible for the generation of these small regulatory RNAs from different precursors. To characterize the cellular site of their generation and accumulation, we purified nuclei from tomato plants infected with potato spindle tuber viroid (PSTVd) RNA, which is known to replicate in the nucleus via double-stranded (ds) RNA intermediates. We could detect PSTVd-specific siRNAs in the cytoplasmic fraction, but not in the nuclear fraction. To correlate the localization of the PSTVd-specific siRNAs with that of similarly sized small RNAs, we studied the compartmentalization of a naturally occurring miRNA. We could detect the precursor of miR167 in the nucleus, but the mature miRNA was found only in the cytoplasmic fraction. We discuss the consequences of this finding for the model of viroid replication and heterochromatin formation.
siRNA and miRNA exert their inhibitory post-transcriptional effect in the cytoplasm, but the effect of siRNAs on heterochromatin formation must occur in the nucleus. This raises the question: in which cellular compartment are the two classes of small RNAs generated; and to what extent do they accumulate in different cellular compartments. It is established that siRNAs are the processing product of perfectly longer dsRNAs that are cleaved by the ribonuclease III-like enzyme Dicer (Bernstein et al., 2001). There is no doubt that generation of siRNA can proceed well in the cytoplasm, as the replicative ds intermediates of plant RNA viruses, including those that are supposed to stay in the cytoplasm during their entire replication cycle, are effectively converted into siRNAs, followed by specific cleavage of single-stranded viral RNA. This process is believed to be the major defense strategy of plants against RNA viruses (Voinnet, 2001). The Dicer enzyme is also responsible for releasing mature miRNAs from nuclear-encoded precursor RNAs that assume a hairpin-type secondary structure (Grishok et al., 2001; Hutvagner et al., 2001). However, different classes of Dicer enzymes are found depending on the organism. While mammals (human, mouse), Caenorhabditis elegans, and the fission yeast Schizosaccharomyces pombe contain only one class of Dicer enzyme, Drosophila melanogaster contains two and there are four in Arabidopsis thaliana (Schauer et al., 2002). Two of those, DICER-LIKE 1 (DCL1) and DCL4, contain nuclear localization signals (NLS). Recently, it was shown that DCL1 is required for the generation of miRNA, but not for the generation of siRNA (Finnegan et al., 2003). In accordance with the NLS signal in DCL1 and based on the expression of nuclear and cytoplasmic variants of P19, which is a suppressor of PTGS derived from tomato bushy stunt virus (TBSV) and known to bind specifically siRNAs, it was recently concluded that plant miRNAs are processed in the nucleus and then exported (Papp et al., 2003). This processing pathway in plants is at variance with the maturation of mammalian miRNAs. In human cell lines, it was shown that the long primary transcripts (pri-miRNA) are processed by the nuclear RNaseIII Drosha into the monomeric stem-loop precursors (pre-miRNA; Lee et al., 2003), which are then exported into the cytoplasm and processed to mature miRNAs (Lee et al., 2002).
To clarify whether in plants siRNAs accumulate in the nucleus, we made use of a viroid of the family Pospiviroidae, which represents a unique case of nuclearly localized dsRNA that is, however, not encoded by the chromosome. Viroids are naked single-stranded, covalently closed circular RNAs that cause infectious diseases in higher plants (Flores, 2001). Their RNA genome does not encode any peptide so that their replication and proliferation requires host factors. Like viruses, they replicate via ds RNA–RNA intermediates; however, the site of replication of Pospiviroidae is confined to the nucleus, where RNA synthesis is carried out by the DNA-dependent RNA polymerase II (Pol II; Schindler and Mühlbach, 1992). Along with others, we could recently show that tomato plants infected with potato spindle tuber viroid (PSTVd) RNA generate viroid-specific siRNAs (Itaya et al., 2001; Papaefthimiou et al., 2001), at first view suggesting that the small RNAs are also formed in the nucleus. According to current models of silencing, nuclear PSTVd-specific siRNA would serve no function. First, they could not direct chromatin changes because so far no homologous DNA sequences have been identified in a host plant. However, if viroid DNA sequences are artificially introduced, they get methylated depending on the presence of replicating viroid RNA (Wassenegger et al., 1994). Second, it is not clear whether a nuclear RISC complex exists (compare Cerutti, 2003) and if so, whether it can direct specific RNA cleavage. In the absence of such a nuclear RISC-driven RNA degradation, viroid-specific siRNAs would not be able to participate in the defense against nuclear viroids. If PSTVd-specific siRNAs would be translocated to the cytoplasm and then incorporated into RISC, they would, according to the model of nuclear viroid replication, not find a specific single-stranded target RNA that they could cleave. Despite the lack of function of PSTVd-specific siRNAs, it should be added that both size classes typical for RNA virus-derived plant siRNA and consisting of about 21/22 and 24–26 nucleotides (Hamilton et al., 2002) could be discriminated. Moreover, viroid-specific siRNAs could also be observed for the peach latent mosaic viroid (PLMVd) and chrysanthemum chlorotic mottle viroid (CChMVd) of the family Avsunviroidae (Martinez de Alba et al., 2002), whose replication and accumulation is restricted to the chloroplast. As there is no evidence for Dicer activity in chloroplasts, those siRNAs must be generated in the cytoplasm. It was our objective to find out where in the plant host cell PSTVd-specific siRNAs actually accumulate. For this purpose, we prepared RNA from purified nuclei originating from PSTVd-infected tomato. We found no experimental indication for viroid-specific siRNA in the nucleus, but they were abundant in the cytoplasmic fraction.
Detection of PSTVd-specific siRNA
We harvested leaves from PSTVd-infected tomato plants and corresponding healthy controls. From each type of plants, a single leaf was taken for the preparation of a total RNA extract, and the residual leaves were used to purify nuclei. Monitoring the preparation under the microscope confirmed: first, the absence of major contamination by chloroplasts; and second, it showed that almost all nuclei were intact. Next, we extracted RNA from the purified nuclei. The four RNA preparations were then subjected to Northern analysis, and initially, we separated the samples on an 8% polyacrylamide gel in order to allow simultaneous detection of all size classes from circular viroid RNA down to siRNAs. Using an RNA probe specific for PSTVd (+) RNA, we could detect circular and linear PSTVd RNAs in the nuclear and in the total RNA fractions (Figure 1a, lanes 2 and 5). The nuclear fraction contained some smaller-than-unit-size RNA fragments, but no siRNAs, which were however clearly visible in the total fraction. The same membrane was then stripped and re-probed for U1 small nuclear (sn) RNA known to occur in both cellular compartments. As expected, U1 could be detected in all RNA fractions (Figure 1b), but the nuclear fractions were distinct as some smaller RNAs could be seen. The membrane was mildly stripped in order to maintain some of the U1 signals and re-hybridized with an RNA probe specific for PSTVd (−) RNA. As anticipated, we found in the total RNA some cross-hybridization with the chloroplast 5S RNA (Figure 1c) as it shares with PSTVd (−) RNA some sequence similarity (Papaefthimiou et al., 2001). We also could detect siRNAs of (−) polarity, however, again only in the total, but not in the nuclear fraction derived from infected plants. Finally, the same membrane was again stringently stripped and re-hybridized with a probe specific for U6 snRNA known to be retained in the nucleus (Boelens et al., 1995; Figure 1d). This experiment essentially repeated the situation as seen for U1 snRNA.
Our finding that PSTVd-specific siRNA was undetectable in the nucleus prompted us to make a more detailed analysis to exclude that this was the result of RNA leakage during the preparation of the nuclei. So far, we could discriminate only between a nuclear RNA fraction and a total RNA fraction, which should contain both nuclear and cytoplasmic RNAs. Therefore, we repeated the preparation of nuclei, but this time we included further cell fractions in the RNA analysis. As described in Experimental procedures, we first homogenized the pool of harvested leaves and kept a part to prepare a total extract (Supplementary Material, Figure S1b). We continued in our preparation of nuclei (Supplementary Material, Figure S1d) and subsequent RNA extraction, but we extracted also RNAs from the remaining fraction, which represents cytoplasmic RNAs (Supplementary Material, Figure S1c) and potentially should contain RNAs that leak out from broken nuclei. To evaluate the quality of this cell fraction, we analyzed the presence of U6 snRNA, but this time separating the RNAs on a denaturing 12% polyacrylamide gel, better suited to resolve small RNAs. Figure 2 shows that the nuclear and total RNA fractions contained U6 snRNA as expected, but that even under overexposure, the cytoplasmic fraction contained only traces of it, demonstrating the efficiency of our fractionation and the quality of our nuclei preparation. It is interesting to note that shorter U6-specific RNA could be found in the nucleus, which could be degradation products, or more likely, they could represent nascent, incompletely synthesized U6 snRNAs of as little as about 40 nucleotides, as purified nuclei are able to continue RNA transcription till they run out on nucleoside triphosphates. There is no indication for leakage of these small U6-specific RNAs into the cytoplasmic fraction. The membrane was then stripped and re-probed for PSTVd (+) and later for PSTVd (−) siRNAs. Consistent with the first experiment, PSTVd-specific siRNA could be detected in the total and also in the cytoplasmic fractions from infected tissues, but not in the nuclear fraction (Figure 2). It is unlikely that leakage of the nuclei is responsible for the identification of siRNAs in the cytoplasmic fraction. If that was the case, also U6-specific smaller RNA should at least, in part, leak out. Moreover, it is hard to envisage that leakage of viroid-specific siRNA would be complete so that even detectable traces would not remain in the nucleus fraction.
Detection of miRNA167 and its precursor
Micro RNAs are generated from a precursor RNA that assumes a hairpin-type secondary structure unlike the complete duplex RNA from which siRNA originates. In accordance with the different substrate RNAs, different enzymes are responsible for the processing reaction, and DCL1 was found associated with miRNA production (Finnegan et al., 2003). We, therefore, also tested the total, cytoplasmic, and nuclear fractions for miR167, which have been detected not only in A. thaliana (Reinhart et al., 2002; Rhoades et al., 2002) but also in tobacco (Mallory et al., 2002), so that there was a good chance that would be present in tomato as well. We used the same membrane as shown in Figure 2 after stripping, and the hybridization with the miR167-specific probe is shown in Figure 3. We could detect the precursor RNA in the nucleus, however, only in the PSTVd-infected nuclei fraction (Figure 3, lane 4). Overlaying the original films of Figures 2 and 3 showed that the position of the pre-miR167 signal was almost identical to that of the U6 snRNA signal (102–104 nts), which is consistent with the length of 101 nts for the precursor of miR167a. We could not detect the precursor in any other fraction, neither in the total fraction nor in the nuclear fraction of the healthy control, possibly because of a hybridization artifact in lane 7. In general, it seems to be difficult to detect pre-miRNAs in plants, and we do not know about any other successful example described. The mature miR167 could be detected in the total and in the cytoplasmic fractions of PSTVd-infected and healthy tomato (Figure 3, lanes 2, 3, 5, and 6) but not in the nuclear fractions. As it was recently concluded that the processing of miRNA proceeds in the nucleus (Papp et al., 2003), this would suggest a rapid exportation of miRNAs, possibly in a concerted reaction following the processing step.
Potato spindle tuber viroid replicates in the nucleus via dsRNA intermediates that are subject to cleavage by one of the Dicer nucleases. Our biochemical analysis has shown that viroid-specific siRNAs are accumulating in the cytoplasm but are not detectable by conventional hybridization methods in the nucleus. Although other small RNAs, such as incomplete fragments of U6 RNA, are retained in the nucleus, possibly because they are associated with protein complexes, it cannot be ruled out that some leakage of siRNAs from the nucleus might occur during the purification procedure. However, regardless of whether more sensitive detection methods might reveal traces of nuclear siRNAs, it is hard to envisage that solely leakage is responsible for the predominantly cytoplasmic accumulation of the PSTVd-specific siRNAs. Although viroids are not derived from nuclear DNA sequences, they do not represent the only example of nuclear dsRNA because duplex RNA can be produced from transposons and artificially introduced chromosomal transgenes. In case of ‘hairpin genes’, there is indirect evidence that a substantial part of the siRNAs accumulate in the cytoplasm. For example, transgenic tobacco generating substantial amounts of hairpin-derived siRNA specific for cucumber mosaic virus (CMV) is resistant to viral infection (Kalantidis et al., 2002), which necessitates a cytoplasmic localization of the siRNAs. There is also evidence for transposon-derived dsRNA in Caenorhabiditis elegans (Sijen and Plasterk, 2003), and for the MuDR/Mu elements that are responsible for the Mutator activity in maize, siRNAs could be detected primarily in the cytoplasmic fraction (Rudenko et al., 2003). However, some short RNAs of antisense polarity could be detected as well in the nuclear fraction, especially in an inactive Mutator line.
The cytoplasmic localization of siRNAs could be the result of two pathways. Either the dsRNAs are exported and then cleaved by Dicer, or alternatively, Dicer cleavage occurs in the nucleus and siRNAs are exported. The latter pathway would be similar to the nuclear processing of plant miRNAs (Papp et al., 2003), which is different from the mammalian pathway (Lee et al., 2002). Assuming that PSTVd-specific siRNA is indeed generated in the nucleus, cleavage should be performed by DCL4, the only other Dicer enzyme with an NLS besides DCL1, which however seems to be exclusively involved in miRNA processing (Finnegan et al., 2003).
Regardless of the site of enzymatic cleavage, the accumulation of PSTVd-specific siRNAs in the cytoplasm is puzzling with regard to their role in viroid replication. PSTVd replicates and accumulates in the nucleus and, in particular, in the nucleolus (Harders et al., 1989), and a recent detailed analysis has shown that the (+) and (−) strands accumulate in different nuclear sites (Qi and Ding, 2003). This results in the paradox situation that the majority of genomic PSTVd RNAs is nuclear, while PSTVd-specific siRNAs are cytoplasmic, leaving them without an actual target for RISC-mediated cleavage. One interpretation of our observation is that unlike what has been assumed so far, a substantial part of either single-stranded and/or ds PSTVd RNAs is actually exported from the nucleus to the cytoplasm, but they are not detectable because the dsRNAs are cleaved by a cytoplasmic Dicer and the single-stranded RNAs of both polarities are cleaved by cytoplasmic RISC, so that they are apparently absent from the cytoplasm. As a consequence of this, the only detectable viroid-specific RNAs in the cytoplasm are siRNAs. This is reminiscent to the observation we described for two transgenic tobacco lines (28 and 65) where no transgene-derived RNA transcript of a ds CMV RNA was detectable, but large amounts of siRNAs (Kalantidis et al., 2002). Further support for RNA turnover in the cytoplasm is provided by Figure 1(a), which shows that PSTVd-specific siRNAs actually represent a substantial portion of the total PSTVd (+) RNA present in an infected cell. Moreover, there is much more PSTVd (−) siRNA than expected, given the low concentration of PSTVd (−) RNA in an infected cell (Spiesmacher et al., 1985). It should be added that PSTVd-specific siRNAs can only be detected if viroid concentration has also reached detectable levels, which is about 1 to 2 days before symptom development (own unpublished results). The presence of cytoplasmic viroid-specific siRNAs (and an activated RISC) would restrict PSTVd and related viroids to the nucleus, a ‘safe’ zone in the cell free of RISC or at least of an activated RISC (for discussion of whether or not there is a nuclear RISC, see Cerutti, 2003). The monomeric covalently closed circular form is not very sensitive to Dicer cleavage (Chang et al., 2003). Because of the circularity and the resulting rod-shaped secondary structure, the monomer may be also less sensitive to RISC-mediated cleavage so that this finally accumulating form of the viroid RNA can pass the cytoplasm. The concept of retreating to a certain cellular compartment would also be in accordance with the occurrence of siRNAs of two chloroplastic viroids (Martinez de Alba et al., 2002). They just prefer another organelle that is not able for RNAi.
An alternative explanation for the occurrence of cytoplasmic viroid-specific siRNA is the possibility that monomeric circular RNA is actually converted in the cytoplasm by RNA-directed RNA polymerase (RdRp) with the aid of siRNAs into cytoplasmic dsRNA, which is then cleaved by Dicer to secondary siRNA. Further, one could envisage that viroid-specific siRNAs are actually generated preferentially in some specialized cell types. Tissue-print hybridizations have shown that PSTVd is preferentially accumulating around the vascular system (Stark-Lorenzen et al., 1997; Zhu et al., 2001). The viroid RNA has to pass this zone for long-distance transport, and thus its titer is high in cells of the vein complex, including companion cells. The mechanisms of transport through the vascular system are supposed to be the same for both families of viroids, regardless of whether they replicate in the nucleus or in the chloroplast. Especially in these cell types, the monomeric circular PSTVd RNA may be a target for RdRp so that both families of viroids will generate siRNAs.
The observation that virally expressed sequences can direct DNA methylation (Jones et al., 2001) indicates that the methylation signal indeed can be imported from the cytoplasm.
Plant infection and growth conditions
Tomato plants (four-leaf stage, cultivar ‘Rentita’) were inoculated with in vitro synthesized RNA transcripts of the severe isolate KF440-2 (Tsagris et al., 1991; Accession number X58388) as previously described by Tabler and Sänger (1984). The plants were kept under greenhouse conditions; they developed PSTVd disease symptoms about 3 weeks after inoculation.
Nuclei purification and cytoplasmic fraction separation
About 6 weeks after inoculation, tomato nuclei were purified from leaves as described by Schumacher et al. (1983), with some modifications. Briefly, 150 g of leaves were immersed in 1.5 l of Buffer A (600 mm mannitol, 25 mm 2-(N-morpholino)ethanesulfonic acid (MES)/NaOH, pH 6.5, 10 mm KCl, 5 mm MgCl2, 0.1% BSA, 1 mm DTT), infiltrated under vacuum, and incubated 30 min on ice. Leaves were then removed from Buffer A, immersed in 450 ml of Buffer B (600 mm sucrose, 25 mm MES/NaOH, pH 6.5, 10 mm KCl, 5 mm MgCl2, 0.1% BSA, 1 mm DTT, 40% glycerol) and homogenized using a blender. The resulting suspension was in part used for RNA extraction (total fraction) or processed further to separate cytoplasmic and nuclear fractions. The filtrated suspension was adjusted to 1% Nonidet P40, and nuclei were sedimented by centrifuging for 10 min at 4000 g. The supernatant was decanted and kept as ‘cytoplasmic fraction’. The pellet, containing the nuclei, was re-suspended in 5 ml of Buffer D (250 mm mannitol, 25 mm MES/NaOH, pH 6.5, 10 mm KCl, 5 mm MgCl2, 0.1% BSA, 1% Nonidet P40, 95% Percoll). After adding 380 ml of Buffer C (600 mm sucrose, 25 mm MES/NaOH, pH 6.5, 10 mm KCl, 5 mm MgCl2, 0.1% BSA, 1% Nonidet P40, 40% glycerol, 1 mm DTT), the suspension was centrifuged for 10 min at 4000 g. The chloroplast-containing dark green layer on the top of the supernatant was sucked off, and the supernatant was discarded. The pellet was re-suspended in 3 ml of Buffer D and 150 ml of Buffer C, and the operation was repeated once. The nuclear pellet was re-suspended in 45 ml of Buffer D and centrifuged for 10 min at 4000 g. The resulting grayish thin layer floating on the top of the suspension, containing the nuclei, was collected, brought to 36% Percoll, and centrifuged for 5 min at 4000 g. The pelleted nuclei were re-suspended in Buffer E (250 mm mannitol, 25 mm MES/NaOH, pH 6.5, 10 mm KCl, 5 mm MgCl2, 0.1% BSA, 1 mm DTT) and stored at −80°C (nuclear fraction).
For this study, we prepared RNA extracts from four different starting materials. Total leaf RNA extracts were prepared as described by Papaefthimiou et al. (2001). In brief, approximately 2–3 g of leaf material was harvested, frozen in liquid nitrogen, and homogenized in a mortar. To the frozen powder, 7 ml of TEMS buffer (100 mm Tris–HCl, pH 7.5, 100 mm NaCl, 10 mm EDTA) supplemented with 100 mm 2-mercaptoethanol prior to usage was added. This was followed immediately by addition of 10 ml of extraction phenol (1 kg phenol, 300 ml of TEMS buffer, 1 g 8-hydroxyquinoline and 250 ml of chloroform). The mixture was vortexed and centrifuged at 4°C, 4000 g for 30 min. The aqueous phase was extracted once more with phenol and then once with chloroform–isoamyl alcohol (24 : 1, v/v), in each case followed by a further centrifugation step. After increasing the sodium concentration to 200 mm by the addition of sodium acetate, pH 5.6, the mixture was precipitated with 2.5 volumes of ethanol. After collecting the precipitate by centrifugation, the samples were washed with 70% ethanol and dried.
For RNA extraction of the three fractions (total, cytoplasmic, and nuclear), the volume was measured and half a volume of TEMS buffer was added, followed by an extraction with an equal volume of chloroform–isoamyl alcohol (24 : 1, v/v). Subsequent phenol extraction and RNA precipitation was as described in the procedure above.
All samples were quantified spectrophotometrically and by loading on ethidium-bromide-stained agarose gels.
Northern blot analysis
Hybridizations were performed using in vitro synthesized 32P-labeled RNA transcripts. The transcripts were generated as previously described using pHa106 as template (Tsagris et al., 1991) linearized with HindIII or EcoRI and transcribed with T7 or SP6 RNA polymerase, respectively, providing longer-than-unit length PSTVd (+) or (−) probes. The control U6 snRNA antisense probe was obtained by transcribing an EcoRI-linearized plasmid containing the mouse U6 snRNA gene with T7 RNA polymerase, and U1 antisense probe (potato) was obtained by transcribing the EcoRI-linearized plasmid pU1EH with SP6 RNA polymerase (Vaux et al., 1992).
Northern blot analysis was performed as described by Papaefthimiou et al. (2001). Briefly, samples equivalent to approximately 100 mg of plant tissue were heat-treated in formamide buffer and loaded onto a 12 or 8% polyacrylamide slab gel (acrylamide, bisacrylamide, 20 : 1) containing 7 m urea and 50 mm TBE buffer (50 mm Tris, 41.5 mm boric acid, 0.5 mm EDTA) and separated by electrophoresis. The samples were electro-blotted to Nytran®N membrane (Schleicher and Schuell, Dassel, Germany) and fixed by UV cross-linking. Pre-hybridization and hybridization, which included the RNA probe at approximately 106 c.p.m. ml−1, was carried out in 5× SSC, 1× Denhardt solution (Sambrook et al., 1989), 1% SDS, 0.25 mg ml−1 tRNA carrier at 58°C for 2 and 16 h, respectively. When the 32P-labeled DNA oligonucleotide was used as probe (Figure 3), the hybridization temperature was 45°C. After hybridization, the filter was washed with 2× SSC, 0.3% SDS for 5 min at room temperature followed by two washes for 20 min at 53°C. In the case of the control hybridization, the membranes were hybridized with 300 000 c.p.m. ml−1 U6 antisense probe in the presence of 50% formamide at 65°C and an additional stringent wash with 0.1× SSC, 0.3% SDS for 15 min at the hybridization temperature was included. Signals were visualized by autoradiography. Stripping of the probe from the membranes was performed by immersing them in boiling buffer (0.1× SSC, 0.5% SDS) for 15 min.
We thank Sergia Tzortzakaki for excellent technical assistance and Drs Kriton Kalantidis and Barthélémy Tournier for critical reading of the manuscript. This work was supported in part by a grant from the General Secretariat for Research and Technology of the Hellenic Ministry of Development (contract PENED 01ED325) and by the European Union (contract QLG2-CT-2002-01673 VIS).