Understanding the relationship between the topological dynamics of nuclear subdomains and their molecular function is a central issue in nucleus biology. Pre-nucleolar bodies (PNBs) are transient nuclear subdomains, which form at telophase and contain nucleolar proteins, snoRNPs and pre-ribosomal RNAs (pre-rRNAs). These structures gradually disappear in early G1 phase and are currently regarded as reservoirs of nucleolar factors that participate to post-mitotic reassembly of the nucleolus. Here, we provide evidence from fluorescence in situ hybridization and loss-of-function experiments in HeLa cells that PNBs are in fact active ribosome factories in which maturation of the pre-rRNAs transiting through mitosis resumes at telophase. We show that the pre-rRNA spacers are sequentially removed in PNBs when cells enter G1 phase, indicating regular pre-rRNA processing as in the nucleolus. Accordingly, blocking pre-rRNA maturation induces accumulation in PNBs of stalled pre-ribosomes characterised by specific pre-rRNAs and pre-ribosomal factors. The presence of pre-ribosomal particles in PNBs is corroborated by observation of these domains by correlative electron tomography. Most importantly, blocking pre-rRNA maturation also prevents the gradual disappearance of PNBs, which persist for several hours in the nucleoplasm. In a revised model, we propose that PNBs are autonomous extra-nucleolar ribosome maturation sites, whose orderly disassembly in G1 phase is driven by the maturation and release of their pre-ribosome content.
During open mitosis, the nucleolus is disassembled like all nuclear domains. Both ribosomal DNA (rDNA) transcription and processing of pre-ribosomal particles are progressively switched off during prophase and then reactivated in telophase (Gébrane-Younès et al., 1997). In prophase, rDNA transcription by RNA polymerase I is inhibited through successive phosphorylation of its co-factors by the cyclin-dependent kinase 1 (CDK1)/cyclinB complex (Heix et al., 1998; Sirri et al., 1999). Numerous components of the transcription machinery, including UBF and most RNA polymerase I subunits, remain associated with the rDNA genes in an inactive form, whereas other proteins transiently dissociate from the transcriptional complexes and relocate to the cytosol (Chen et al., 2005; Leung et al., 2004; Roussel et al., 1996; Sirri et al., 1999; Zatsepina et al., 1993). Although the pre-rRNA processing machinery is inhibited during mitosis, stable pre-rRNAs are detected in mitotic cells (Beven et al., 1996; Dousset et al., 2000; Dundr and Olson, 1998; Dundr et al., 2000; Fan and Penman, 1971; Piñol-Roma, 1999). When transcription is shut down, these pre-RNAs progressively diffuse from the nucleoli to the cytosol and form a sheath around the condensed chromosomes in metaphase; they are also detected in a few cytosolic foci called nucleolar-derived foci (NDF), which persist in the cytosol of daughter cells until early G1 (Dundr and Olson, 1998; Dundr et al., 2000). In telophase, the transcriptional machinery is reactivated at the nucleolar organiser regions (i.e. the active rDNA loci), where nucleoli reform. As the chromosomes start to unwind, the perichromosomal sheath breaks down into pre-nucleolar bodies (PNBs), which contain pre-rRNAs, but not the RNA polymerase I machinery (Chan et al., 1991; Dundr and Olson, 1998; Dundr et al., 2000; Fomproix and Hernandez-Verdun, 1999; Jimenez-Garcia et al., 1994; Roussel et al., 1993; Savino et al., 1999). Some PNBs fuse with the reforming nucleoli at telophase/early G1, while others remain in the nucleoplasm and progressively resorb in G1.
Based on their content and on their morphology, PNBs have long been considered as intermediates in the reconstitution of the incipient nucleoli. More recently, it has been proposed that pre-rRNAs present in PNBs might act as assembly sites, complexing processing factors before their incorporation into the nucleolus (Angelier et al., 2005; Dousset et al., 2000; Dundr et al., 2000; Muro et al., 2010; Savino et al., 2001). According to this hypothesis, the stability of the pre-rRNAs is the key factor that governs retention of nucleolar proteins in PNBs. However, the underlying determinants of pre-rRNA stability have remained vague. In addition, direct evidence in support of a role of PNBs in nucleolus re-assembly is still lacking.
In eukaryotes, rDNA transcription in the nucleolus produces a large polycistronic precursor, called 47S pre-rRNA in vertebrates. This early pre-rRNA contains the sequences of the 18S, 5.8S and 28S rRNAs, flanked by external and internal transcribed spacers (ETSs and ITSs; Fig. 1). A subset of ribosomal proteins and pre-ribosomal factors assemble co-transcriptionally with the 47S pre-rRNA, forming the 90S pre-ribosomal particle (Henras et al., 2008; Pérez-Fernández et al., 2007). Pre-ribosomal particles are then further matured by the sequential removal of the ETS and ITS sequences, through endonucleolytic cleavages or exonucleolytic processing (Borovjagin and Gerbi, 2005; Eichler and Craig, 1994; Hadjiolov, 1985). Cleavage of the ITS1 at site 2 (Fig. 1) splits the 90S particle into pre-40S and pre-60S ribosomal particles, which then follow distinct maturation pathways. The final processing steps take place after export into the cytoplasm, leading to the two mature ribosomal subunits (Rouquette et al., 2005; Thomson and Tollervey, 2010). Pre-rRNA maturation involves almost 200 trans-acting proteins and over 100 small nucleolar RNPs (snoRNPs) that intervene sequentially in an intricate ballet of assembly and disassembly steps, as shown in the yeast Saccharomyces cerevisiae (Fatica and Tollervey, 2002; Henras et al., 2008; Kressler et al., 2010).
Current knowledge regarding the fate of the pre-rRNAs in PNBs is fragmentary. Fluorescence in situ hybridization (FISH) experiments have shown that PNBs contain partially processed pre-rRNAs lacking the 5′-ETS leader sequence and the 3′-ETS (Dundr and Olson, 1998; Dundr et al., 2000), two sequences that are removed early in the processing pathway. In addition, the 5′-ETS was detected in PNBs at telophase, but not in G1 phase. A number of trans-acting factors have been localised in PNBs, together with pre-rRNAs, including the small nucleolar RNAs (snoRNAs) U3 and U14 and some pre-ribosomal proteins (for a review, see DiMario, 2004). Strikingly, these factors exit from the PNBs in an ordered way that correlates with the order in which they participate in pre-rRNA processing. Proteins required for the earliest steps of ribosome biogenesis, such as fibrillarin, leave PNBs first, while pre-ribosomal factors (e.g. Bop1, Nop52 or B23), which participate in subsequent steps, reside in PNBs for a longer period of time (Angelier et al., 2005; Dundr et al., 2000; Leung et al., 2004; Savino et al., 1999; Savino et al., 2001). Pre-ribosomal factors in PNBs are rapidly exchanged with the rest of the nucleus and can establish functional interactions in PNBs, as shown for Nop52 and B23 (Muro et al., 2010).
Here we have investigated whether pre-rRNAs are subjected to maturation in PNBs, like pre-rRNAs in the nucleolus. Consistent with this idea, PNBs were previously shown to contain PM-Scl100, a component of the exosome, a multi-enzymatic exonucleolytic complex involved in pre-rRNA processing (Fomproix and Hernandez-Verdun, 1999). We have found that this hypothesis is fully supported by the evolution of pre-rRNAs in PNBs from mitosis to G1, by loss-of-function experiments targeting pre-rRNA processing, as well as by the observation of the subcellular structure of PNBs by electron tomography. These results offer a parsimonious explanation for the ordered release of pre-ribosomal factors from PNBs as pre-ribosome maturation progresses. In the light of these data, we argue that PNBs are primarily maturation sites for mitotic pre-rRNAs.
Pre-rRNAs in PNBs evolve from early to late precursors
To determine whether pre-rRNAs were processed in PNBs, we looked at their evolution from mitosis to G1. FISH was carried out with a combination of seven probes complementary to different segments of the transcribed spacers to discriminate early versus late precursors (Fig. 1). This experiment was performed on asynchronous HeLa cells expressing B23/nucleophosmin fused to GFP (HeLa GFP-B23), a well-described marker of PNBs (Louvet et al., 2005). Labelling by the different probes of NDFs and PNBs, detected with GFP-B23, was scored in cells from anaphase to G1 phase (Fig. 2).
In mitotic cells, from prometaphase to telophase, probes complementary to the 5′-ETS, ITS1 or ITS2 labelled both the chromosomal sheath and the NDFs (Fig. 2A,B). No labelling was observed in these structures with probes specific either for the 5′-ETS leader sequence (ETS1-297) or for the 3′-ETS (probe 28S-3′ETS), indicating that mitotic pre-rRNAs had already undergone the earliest cleavages at site A′ and at the end of the 28S segment (data not shown). Cells in late anaphase/telophase displayed NDFs and PNBs labelled with GFP-B23. At this stage, these structures were all labelled with the 5′-ETS-, ITS1- and ITS2-specific probes (Fig. 2B,C). In contrast, only 20% of the PNBs contained 5′-ETS in early G1 cells, indicating the loss of early rRNA precursors. Furthermore, while the ITS2-specific probes labelled all the PNBs, the ITS1 was only detected in 80% of them (Fig. 2B,C). In cells further engaged in G1, displaying smaller and fewer PNBs in rounder and larger nuclei, PNBs were only labelled with ITS2-specific probes, indicating the sole presence of 60S subunit precursors (Fig. 2B,C). We never detected PNBs containing ITS1 without ITS2, or 5′-ETS without ITS1 and ITS2, indicating that the presence of pre-rRNAs in PNBs is not random, but rather follows a precise timing (Fig. 2C): the 5′-ETS is first to disappear at the end of telophase, followed by the ITS1 in early G1 and ultimately by the ITS2, which corresponds exactly to elimination of the transcribed spacers during pre-rRNA processing. Importantly, PNBs remained negative upon staining of the cells for 30 minutes by ethynyl-uridine (EU), indicating that no neosynthesised pre-rRNAs transited through these structures (supplementary material Fig. S1).
Specific pre-rRNAs accumulate in PNBs upon ribosomal protein knockdown
These observations were fully consistent with pre-rRNA maturation taking place in PNBs. To further challenge this hypothesis on mechanistic grounds, we checked whether altering pre-rRNA processing could specifically modulate the fate of pre-rRNAs in PNBs. We knocked down a series of ribosomal proteins involved in the biogenesis of the small or the large subunits, from early to late stages. Upon knockdown of the small ribosomal subunit protein RPS29, which affects the maturation of the 21S and 18S-E pre-rRNAs (O'Donohue et al., 2010), PNBs in cells exiting mitosis contained ITS1 and ITS2 (supplementary material Fig. S2). The 5′-ETS disappeared at telophase, as observed in control cells (data not shown). However, at later stages in G1, we observed the persistence of the ITS1 in PNBs that did not contain ITS2 anymore, consistent with the accumulation of 21S and 18S-E pre-rRNAs observed on northern blot (Fig. 3A). In contrast, depletion of the large ribosomal subunit protein RPL5 caused accumulation of PNBs labelled with the ITS2-specific probe, but not with the ITS1-specific probes (Fig. 3A). When proteins necessary for processing of the 5′-ETS were depleted, like RPS5, RPS6 or RPS24, this transcribed spacer remained visible in PNBs of G1 cells, as illustrated for RPS5 in Fig. 3B; these late PNBs were also labelled both with the ITS1- and the ITS2-specific probes, consistent with accumulation of 30S and 45S pre-rRNAs (Fig. 3A). An intermediate situation was observed when RPS19 was knocked down: late PNBs contained the ITS1 and ITS2 (Fig. 3A), but not the 5′-ETS, as expected from the accumulation of 21S and 41S pre-rRNAs observed in RPS19-depleted cells (Fig. 3B). This observation was not limited to ribosomal proteins since similar results were observed upon knockdown of bystin/ENP1, a pre-ribosomal factor required for the same maturation steps as RPS19 (Fig. 3B).
Pre-rRNA maturation defects block PNB disappearance in G1 phase
In addition to changing the content in pre-rRNAs of PNBs, perturbing pre-rRNA maturation also resulted in a significant increase of the proportion of interphasic cells containing PNBs (30–50%) as compared to control cells (10%), suggesting a difference in the lifetime of PNBs (see Fig. 5). This led us to assess the dynamics of PNBs in cells depleted of RPS19 or bystin by performing time-lapse fluorescence microscopy with GFP-B23, from anaphase to G1. The kinetics of the formation of PNBs at the end of mitosis was similar to control cells: a fraction of them fused with the incipient nucleoli, while the others remained isolated in the nucleoplasm (illustrated for RPS19 depletion in supplementary material Fig. S3; Movie 1). But in contrast to control cells, in which PNBs disappeared within 2–3 hours, a large number of PNBs persisted for several hours in the nucleoplasm of cells depleted of RPS19 or bystin (illustrated for bystin in Fig. 4; supplementary material Movie 2). Therefore, altering pre-rRNA processing not only changed the pre-rRNA composition of late PNBs, but also strongly delayed the progressive dissolution of PNBs in G1. These data indicate that PNB disappearance in G1 depends on the processing of the pre-rRNAs that they contain. By labelling the cells with 5-ethynyl-2′-deoxyuridine (EdU), we verified that the persistent PNBs observed in RPS19-depleted cells were present in G1 phase, but not in S and G2 phase (Fig. 5). As reported before (Miyake et al., 2008), we also observed that depletion of RPS19 induced an arrest or a strong delay of the cells in G1 phase (not shown).
Specific pre-ribosomal particles accumulate in stalled PNBs
The pre-ribosomes accumulating in PNBs upon alteration of pre-rRNA processing were further characterised by detecting some pre-ribosomal factors, either with specific antibodies or by expressing fluorescent fusion proteins. Upon knockdown of RPSs, proteins hDim2, bystin/hEnp1 and hNoc4, three components of the 90S and pre-40S particles, accumulated in PNBs together with the U3 snoRNA, which is involved in several steps of the small subunit maturation pathway (Borovjagin and Gerbi, 2005) [illustrated in Fig. 6A,B for rps19 short interfering RNA (siRNA)-treated cells]. In contrast, none of these components was detected in PNBs persisting after depletion of RPL5, as illustrated for hNoc4 and U3 (Fig. 6B). They contained in turn the pre-60S ribosomal factors Bop1 (Lapik et al., 2004) and Nop52 (Savino et al., 1999), as well as the large ribosomal subunit protein RPL29 (Fig. 6C). Neither Bop1, nor Nop52, nor RPL29 were present in PNBs persisting upon knockdown of RPSs or bystin (Fig. 6C; supplementary material Fig. S2 for RPS29 knockdown). NPM/B23 was always observed in persistent PNBs, irrespective of the siRNA used, in agreement with the versatile function of this protein in the production of both ribosomal subunits (Lindström, 2011). The snoRNP protein fibrillarin and the U8 snoRNA, which participate in early maturation steps, were always absent from late PNBs (data not shown).
These results are fully consistent with the presence of partially assembled pre-ribosomal particles in PNBs upon interruption of the maturation process. Interestingly, late PNBs accumulating upon RPSs knockdown contained neither the late pre-40S particle components hLtv1 and hRio2, which are presumably involved in nuclear export and final cytoplasmic processing of the pre-40S particles (Rouquette et al., 2005; Wild et al., 2010; Zemp et al., 2009), nor hNob1, the putative endonuclease ensuring the ultimate cleavage of the 3′ end of the 18S rRNA in the cytoplasm (Lamanna and Karbstein, 2009) (Fig. 6A). These late pre-ribosomal factors were not detected in nucleoli either (Fig. 6A), unlike hDim2, bystin and Noc4, which suggests that the presence of pre-ribosomal factors in PNBs obeys the same laws as in nucleoli.
Correlative electron tomography shows pre-ribosomal particles in PNBs
Finally, we sought to directly visualise the pre-ribosomal particles in PNBs by correlative electron tomography in HeLa GFP-B23 cells. GFP-labelled PNBs were first detected in ultrathin sections by fluorescence microscopy before observation by transmission electron microscopy (Fig. 7A). In conventional electron micrographs, PNBs appeared as dense fibrillo-granular bodies, often located in the vicinity of the nuclear envelope (Fig. 7B,C), similar to previous descriptions (Azum-Gélade et al., 1994; Dundr et al., 2000; Savino et al., 2001). When these structures were imaged by electron tomography, isolated particles became more clearly distinguishable in PNBs (Fig. 7D) and could be segmented in three dimensions (Fig. 7E,F). For a large fraction of the particles, the size of the bounding boxes (minimal bounding rectangles) framing their median section in the tomogram was in the same range as projections of randomly oriented 40S and 60S ribosomal subunits, arguing for these particles being medium or late pre-ribosomal intermediates (Fig. 7G). In addition, larger complexes (up to 30 nm in length) were observed, potentially corresponding to early pre-ribosomes. Indeed, the size range of the particles in PNBs of RPS19-depleted cells was narrower than in control cells: it matched the computed dimensions of the 40S subunit (Fig. 7G; F-test: P>0.15 for both dimensions), consistent with the accumulation of late nuclear pre-40S particles in these structures. These structural data are in good agreement with the molecular characterisation of PNBs presented above. Taken together, these results strongly support that mitotic pre-rRNAs gathering in PNBs at the end of mitosis undergo maturation, which drives the dynamics of PNBs up to their disappearance in G1 phase.
The data presented here strongly converge to show that mitotic pre-rRNAs undergo processing in PNBs. FISH detection of pre-rRNAs using a broad range of probes established that pre-rRNAs within PNBs evolve during the course of the G1 phase, following the same order as processing in nucleoli. More precisely, the disappearance of the 5′-ETS at the telophase/G1 transition is followed by the progressive removal of ITS1 and ITS2, the latter being more persistent within PNBs. This persistency of the 60S subunit precursors in PNBs corresponds to the longer maturation time of these subunits, as compared to the 40S subunits. Consistently, it was shown in monkey CMT3 that 18S rRNA precursors disappeared from PNBs after mitosis more rapidly than 28S rRNA precursors (Dundr et al., 2000). Maturation of pre-rRNAs in PNBs was further demonstrated by loss-of-function experiments, since knockdown of ribosomal proteins or pre-ribosomal factors blocked the maturation of pre-rRNAs in PNBs. Specific pre-rRNAs accumulated in PNBs depending on the maturation stage affected by protein knockdown, together with specific nucleolar factors (pre-ribosomal proteins, snoRNAs), which indicates that these domains contain pre-ribosomal particles whose maturation was interrupted. Indeed, direct visualisation of PNBs by electron tomography revealed the presence of putative pre-ribosomes, whose size was compatible with the accumulation of pre-40S particles in RPS19-depleted cells.
In the light of the present data, PNBs emerge as autonomous extra-nucleolar sites in which processing of mitotic pre-rRNAs takes place. This description of PNBs is compatible with the dynamic nature of these structures, which allows for the rapid exchange of pre-ribosomal factors, as observed for fibrillarin, Nop52 and B23 using FRAP and photo-activation experiments (Dundr et al., 2000; Muro et al., 2010). In addition, Nop52 and B23, which are involved in ITS2 processing, are dissociated in the mitotic chromosomal sheath, but do form a complex in PNBs, indicating that pre-ribosomal factors establish functional interactions in PNBs (Angelier et al., 2005). The presence of pre-ribosomal factors in these foci appears to follow the same pattern as in nucleoli, as shown here for the early and late factors of the 40S maturation pathway: the late factors hRio2 and hLtv1, which are not found in nucleoli, do not accumulate in PNBs either, while bystin/hEnp1, hDim2 and hNoc4 are found both in nucleoli and in PNBs. As described for nucleoli, the structure of PNBs may be maintained through self-organisation of the processing complexes around pre-rRNAs. This interpretation of PNB structure is supported by recent data that showed that several types of coding or non-coding RNAs can nucleate the formation of nuclear bodies (Mao et al., 2011; Shevtsov and Dundr, 2011). However, unlike nucleoli, the position of PNBs in the nucleus is not constrained by rDNA chromosomal loci (NORs). In this respect, it is possible to draw a parallel between PNBs and the small autonomous nucleolar domains (mini-nucleoli) that form when ribosomal genes are spread on individual plasmids in S. cerevisiae (Oakes et al., 1998; Trumtel et al., 2000).
In reforming nucleoli, recruitment of the nucleolar components follows the order of pre-rRNA maturation and is presumably dictated by the presence of pre-rRNAs. Our data strongly argue that, conversely, the driving force behind the disassembly of PNBs is the progressive release of pre-ribosomal factors whose role in pre-ribosome maturation has been fulfilled. In this model (Fig. 8), the release from PNBs of early acting factors takes place at the same time as the processing of the most immature pre-rRNA species at telophase (Fig. 8A). Afterwards, intermediate processing factors involved in the 40S subunit maturation pathway are released next. Some of these are exported to the cytoplasm together with pre-40S particles. Ultimately, PNBs only contain pre-60S particles, which are progressively exported from the nucleus. As predicted from this model, blocking pre-rRNA maturation is sufficient to prevent disassembly of PNBs. Hence, when RPS19 or bystin/hEnp1 are depleted (Fig. 8B), pre-60S particles are matured in PNBs and exported to the cytoplasm, but the pre-40S particles accumulate, thus blocking resorption of PNBs.
Although our current description of PNBs does not contradict previous data, it does call for a partial re-interpretation of some results. While they have mostly been envisioned as domains involved in post-mitotic reformation of nucleoli, PNBs appear to be dedicated to mitotic pre-rRNA processing, in parallel with, but independently of, the processing of neo-synthesised pre-rRNAs in the incipient nucleoli. To our knowledge, data showing a direct causal link between the dynamics of PNBs and the reformation of nucleoli at the end of mitosis are still to be established. The maintaining of assembled pre-ribosomes through mitosis is indeed likely to promote priming of the pre-rRNA processing machinery in daughter cells. In addition, several signalling pathways have now been shown to induce cell cycle arrest in interphasic cells in response to ribosome biogenesis defects (Boulon et al., 2010). One may speculate that resumption of intermediate and late steps of pre-rRNA processing immediately at the end of mitosis in PNBs, including those that fuse with incipient nucleoli, prevents untimely activation of these ribosomal stress responses during reformation of nucleoli. Noticeably, haplo-insufficiency of several ribosomal proteins, including RPS19, RPS24 and RPL5, is responsible for Diamond-Blackfan anemia, a congenital erythroblastopenia (Ellis and Gleizes, 2011), which could correspond to a physiological situation where defects in PNB dynamics occur. Finally, as autonomous pre-rRNA processing domains, PNBs could prove to be a good experimental model to analyse the formation of pre-ribosomes in situ and study nucleolar dysfunctions, as initiated here by imaging pre-ribosomes directly in PNBs using correlative electron tomography.
Materials and Methods
Cell culture and reagents
Human cervical carcinoma HeLa cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS) and 1 mM sodium pyruvate, in 5% CO2 at 37°C. HeLa cells, stably expressing EGFP fused to the N-terminus of the nucleolar factor B23 (HeLa GFP-B23 cells) or EYFP fused to the C-terminus of RPL29 (HeLa RPL29-YFP), were kindly provided by Drs D. Hernandez-Verdun (Louvet et al., 2005) and U. Kutay (Wild et al., 2010), respectively. Transient expression of GFP-Noc4 (cloned in the HindIII-XbaI sites of pEGFP-C3), GFP-Bop1 (cloned in the EcoRI site of pEGFP-C2, a gift from Dr D. Pestov) or GFP-Nop52 (cloned in the EcoRI site of pEGFP-C2, a gift from Dr P. Roussel) was performed by electrotransformation of 1 µg plasmid/106 cells, as described in the next section.
Knockdown of protein expression
Different 21-mer siRNAs were used to knock down expression of the following human proteins: for RPS5 (GenBank accession no. NM_001009), 5′-CCGAUGAUGUGCAGAUCAAdTdT-3′ (rps5 siRNA); for RPS6 (GenBank accession no. NM_001010), 5′-GACGCAGACUUUCCUCUCUdTdT-3′ (rps6 siRNA); for RPS19 (GenBank accession no. NM_001022), 5′-AGAGCUUGCUCCCUACGAUdTdT-3′ (rps19 siRNA); for RPS24 (GenBank accession no. NM_033022), 5′-CACCGGAUGUCAUCUUUGUdTdT-3′ (rps24 siRNA); for RPS29 (GenBank accession no. NM_001032), 5′-CGGUCUGAUCCGGAAAUAUdTdT-3′ (rps29 siRNA); for RPL5 (GenBank accession no. NM_000969), 5′-GCCACAAUGUUGCAGAUUAdTdT-3′ (rpl5 siRNA); for Bystin (GenBank accession no. NM_004053), 5′-CUGCCCAAGGCAUUUAAGAdTdT-3′ (bystin-1 siRNA) and 5′-GAGGUUGAGACAGUCAUGUdTdT-3′ (bystin-2 siRNA). All siRNAs were purchased from Eurogentec and provided as duplexes diluted 100 µM. 10 µl of siRNA solution were added to 107 HeLa cells in suspension in 200 µl of DMEM without FCS. For bystin knockdown, a mixture of 10 µl each of bystin-1 and bystin-2 siRNAs were used. For controls, cells were electro-transformed either with a scramble siRNA or without siRNA, which lead to similar results. Electrotransformation was performed at 250 V and 970 µF with a Gene Pulser (Bio-Rad Laboratories), in a cuvette with a 4-mm inter-electrode distance. The cells were diluted in 20 ml DMEM supplemented with FCS and antibiotics and plated in a 140-cm2 Petri dish for RNA extraction and northern blot analyses. They were also seeded either on glass coverslips laid in 24-well microplates for FISH and immunofluorescence assays, or on glass-bottomed Petri dishes 35 mm in diameter (Dominique Dutscher) for live-cell imaging.
Northern blot analysis of pre-rRNA
Total RNAs were purified with Trizol reagent. After alcohol precipitation, RNA pellets were dissolved in formamide, quantified using a Nanodrop spectrophotometer (Thermo Fisher Scientific) and diluted to 1 mg/ml. Pre-rRNA analysis by northern blot was performed as described previously (O'Donohue et al., 2010) with the following probes: 18S (5′-TTTACTTCCTCTAGATAGTCAAGTTCGACC-3′), 28S (5′-CCCGTTCCCTTGGCTGTGGTTTCGCTAGATA-3′), 5′-ITS1 (5′-CCTCGCCCTCCGGGCTCCGTTAATGATC-3′), ITS2-b (5′-CTGCGAGGGAACCCCCAGCCGCGCA-3′), and ITS2-d/e (5′-GCGCGACGGCGGACGACACCGCGGCGTC-3′). For detection of the ITS2, the ITS2-b and ITS2-d/e probes were mixed in equal amounts. After hybridisation, the membranes were washed twice for 10 min in 2× SSC, SDS 0.1%, and once in 1× SSC, SDS 0.1%. Labelled RNA signals were acquired with a phosporimager (FLA2000; Fuji) and quantified with the Image Gauge software.
Labelling of cellular DNA with EdU
Cells were incubated with 20 µM 5-ethynyl-2′-deoxyuridine (EdU, Invitrogen), an analogue of thymidine (Salic and Mitchison, 2008), in culture medium for 4 h. At the end of the reaction time (which corresponded to 48 h post-transfection), the cells were rinsed twice in PBS and fixed with 4% paraformaldehyde in PBS for 15 min. After several washes in PBS, cells were permeabilised for 5 min in 0.5% Triton X-100 diluted in PBS. They were then rinsed twice in PBS containing 3% BSA. EdU incorporated into DNA was revealed by click chemistry through the reaction of its terminal alkyne group with an azide derivative coupled to Alexa-594 (Invitrogen). The reaction was developed for 30 min in the presence of TBS buffer (150 mM NaCl, 50 mM Tris-HCl, pH 8.0) containing 4 mM CuSO4, 10 µM Alexa-594 azide, 2 mg/ml ascorbic acid. The coverslips were protected from light and incubated at ambient temperature under slow agitation. The cells were then rinsed briefly in TBS buffer. PNBs were revealed by FISH with the 5′ITS1 probe conjugated to Alexa-488 (O'Donohue et al., 2010). After two washes in 2× SSC containing 10% formamide, DNA was counter-stained with Hoechst 33342 (Invitrogen) and the coverslips were mounted in Mowiol (version 4.88; PolyScience, Inc.).
Fluorescence in situ hybridisation and immunofluorescence assays
Pre-rRNA species were localised by fluorescence in situ hybridisation (FISH) as described (O'Donohue et al., 2010). Appropriate combinations of the following probes, conjugated to Cy3, Cy5 or Alexa-488, were chosen in order to reveal various subsets of pre-rRNA species: ETS1-297 (5′-AGACGAGAACGCCTGACACGCACGGCAC-3′), ETS1-1399 (5′-CGCTAGAGAAGGCTTTTCTC-3′), ETS1-18S (5′-GATCAACCAGGTAGGTAAGGTAGAGCGCGGCGA-3′), 5′ITS1 (5′-CCTCGCCCTCCGGGCTCCGTTAATGATC-3′), ITS1-160 (5′-TCTCCCTCCCGAGTTCTCGGCTCT-3′), 5.8S-ITS2 (5′-GCGATTGATCGGCAAGCGACGCTC-3′), 28S-ETS2 (5′-CGTGCGTGCGCGCGCGCCTGTTTGGGAAC-3′). For snoRNAs, probes rU3 (5′-TTGCACAGAAGCAGCACCTAGAGCCGGCTTCACGCTTT-3′) and hU8 (5′-ATAGGAGCAATCAGGATGTTGCAAGACCTGATTACGCAGAGACGTT-3′) conjugated in their 5′ end to Cy3 were used. Cells grown on glass coverslips were fixed with 4% paraformaldehyde in PBS for 30 min. After washing in PBS, the cells were permeabilised in 70% ethanol overnight at 4°C. They were then re-hydrated twice for 5 min in 2× SSC containing 10% formamide. Hybridisations in the presence of two to three different probes and further processing of the coverslips were as described in the previous section.
For immunofluorescence studies, cells grown on glass coverslips were fixed for 15 min with 4% paraformaldehyde in PBS. After washing in PBS, cells were permeabilised for 5 min in 0.1% Triton X-100 and 0.02% SDS in PBS. Saturation was performed for 30 min first in 2% BSA in PBS, and then in 10% normal goat serum (NGS) in PBS. The primary antibody, diluted in 10% NGS in PBS, was put in contact with the cells for 1 h at room temperature. The antisera were: anti-Rio2 (1∶5000), anti-Nob1 (1∶5000), anti-Dim2 (1∶2000) [all gifts from Dr U. Kutay (Zemp et al., 2009)], [anti-bystin (1∶10,000), anti-Ltv1 (1∶20,000); both raised in our laboratory], and anti-fibrillarin (1∶1000; a gift from Dr J. Cavaillé). After washing in 2% BSA in PBS, and in 10% NGS in PBS, cells were incubated with the secondary antibody coupled to Alexa-488, 555 or 647 and diluted 1∶4000 in 10% NGS in PBS. After washing briefly in 0.1% Triton X-100 and 0.02% SDS in PBS, DNA counterstaining with Hoechst and mounting were as described above.
Light microscopy observations and time-lapse experiments
Observations of fixed cells were made with an inverted microscope (IX-81; Olympus), equipped with a motorized stage (Märzhäuser) and a camera (Coolsnap HQ; Photometrics), driven by MetaMorph (MDS Analytical Technologies). Cells were observed using a ×100, oil immersion objective (Plan Apochromat, 1.4 NA; Olympus) driven by a piezo z-stepper (PI). The pixel size was 64.5 nm. The light source was a monochromator (Polychrome V; Till Photonics GmbH) equipped with a 150 W Xenon source and used with a 15-nm bandwidth. For multi-labelling experiments, multiband dichroic mirrors (Chroma) were used to avoid pixel shift in the acquisition plane. Specific single-band emission filters were mounted on a motorized wheel (PI). To analyse EdU incorporation experiments by image cytometry, adjacent fields were systematically acquired in each experimental condition, until an appropriate number of cells had been reached (between 700 and 1100 cells). By using Hoechst and 5′ITS1 labelling, each nucleus was then checked for its content in PNBs. In each group, EdU-labelled versus unlabelled cells were assessed.
Time-lapse experiments were done at 37°C on live HeLa GFP-B23 cells 48 h post-transfection with siRNAs. Images were acquired with the wide-field microscope described above with a 0.3 µm z-step. Alternatively, 3D time-lapse experiments were performed with an Andor Revolution Nipkow-disk confocal system installed on an inverted microscope (IX-81; Olympus), featuring a confocal spinning disk unit (CSU22; Yokogawa) and a cooled EM-CCD camera (Andor). For illumination, the 488-nm line of a diode pumped solid-state laser (25 mW; Coherent) was used at 30% of its power. With a 100× oil immersion objective (Plan Apochromat, 1.4 NA; Olympus), the pixel size was 77 nm. The exposure was set to 100 ms per plane. For comparison purposes, cells in anaphase or very early telophase were selected, as the former is the shortest step of mitosis. Images were captured every 3 min over 2 h in order to follow the last steps of mitosis and early G1. Alternatively, they were acquired every 10 min for the first 3 h, and then every 30 min for up to 8 h in order to study the behaviour of residual PNBs during G1 phase in siRNA-treated cells. Deconvolution, maximum intensity projections and image analyses were performed with MetaMorph software (Molecular Devices).
Correlative electron tomography
HeLa GFP-B23 cells were prepared for TEM by overnight fixation at 4°C with 3% paraformaldehyde and 0.1% glutaraldehyde in phosphate buffer (pH 7.2). They were collected by scrapping, centrifuged briefly and the pellet was included in agarose. Embedding in resin was performed according to the Progressive Lowering of Temperature method using an automatic freeze-substitution system (EM AFS2; Leica Microsystems). The cells were cryoprotected with 3.5% sucrose and stained with 1% uranyl acetate. After dehydrating with ethanol while decreasing the temperature from 0°C to −30°C, the samples were infiltrated with Lowicryl HM20 resin which was polymerized under UV light for 2 days at −30°C. To perform correlative microscopy, ultra-thin sections were cut on a UC6 ultramicrotome (Leica Microsystems) at room temperature with a nominal thickness of 150 nm using an ultra 45° diamond knife (Diatome Inc.) and deposited on formvar-carbon-coated Finder grids (EMS). The grids were mounted on a glass slide in 50% glycerol and observed by fluorescence microscopy to detect PNBs with GFP-B23. The alphanumerical code of the Finder grids was used to position the cells of interest. The grids were then removed from the glass slides, extensively washed in water to remove glycerol, stained with uranyl acetate and lead citrate and 10 nm-diameter colloidal gold particles were deposited onto both sides. Electron tomography was performed on the same sections by recording tilt-series at 200 kV with a LaB6 TEM (JEM-2100; Jeol) equipped with a 2×2 k US1000 CCD camera and driven by Digital Micrograph software (Gatan Inc.). The angular range was −60° to 60° sampled with 2° increments. The nominal magnification was 15,000×, which resulted in a pixel size of 6.9 Å at the specimen level. Tomographic reconstructions were performed with the simultaneous iterative reconstruction technique provided by IMOD software (Kremer et al., 1996) using the 10 nm gold particles as fiducial markers for aligning the image stack. Particles were visualised with the University of California San Francisco Chimera package (Pettersen et al., 2004) using the Segment Map function from the Segger package (Pintilie et al., 2010). For measurement of their size, particles were manually extracted from the reconstructed tomograms and a bounding box was adjusted around the middle section of each particle. The longest and shortest dimensions were attributed to the length and width, respectively. These bounding boxes were compared with those obtained from the X-ray crystal structures of the yeast 40S and 60S ribosomal subunits [Protein Data Bank accessing codes 3O2Z and 3O58, respectively (Ben-Shem et al., 2010)]. In order to do so, the structures were filtered at 80 Å resolution and converted to mrc format with a 6.9 Å per pixel step, using Chimera. Projections of the filtered 3D models were then computed using IMAGIC-5 (van Heel et al., 1996). For each 3D model, a set of 108 projections uniformly distributed within the asymmetric triangle was produced, using an angular increment of 20° along the longitudinal axis. The projections were then normalised and binarised with a 1σ threshold and the dimensions of the bounding boxes fitting the projection were measured with IMOD as described above. Electron tomography was performed at the electron microscopy facility of the Fédération de Recherche en Biologie de Toulouse.
The authors are grateful to Drs Ulrike Kutay for the gift of antisera and of the HeLa RPL29-YFP cell line; Danièle Hernandez-Verdun and Pascal Roussel for the HeLa GFP-B23 cell line and the pEGFP-Nop52 construct; Dimitri Pestov for pEGFP-Bop1; Céline Verheggen and Jérôme Cavaillé for snoRNA probes; and Christophe Normand for his help with the spinning disk confocal microscope. The authors also wish to thank Dr C. Verheggen for her early interest in this study and helpful advice, Valérie Choesmel for her involvement at the beginning of the project and Drs Jérôme Cavaillé, François Morlé and Boris Guyot for scientific discussions. Pierre-Emmanuel Gleizes and Marie-Françoise O’Donohue contributed equally to this work as senior authors.
This project was funded by the Centre National de la Recherche Scientifique (CNRS); the University of Toulouse, the French National Research Agency [grant number ANR-06-BLAN-0037 to P.E.G.]; the Association pour la Recherche contre le Cancer [grant number 1058 to P.E.G.]; and the Institut Universitaire de France to P.E.G. C.C. was supported by fellowships from the Ministère de l’Éducation Nationale et de la Recherche and the Association pour la Recherche sur le Cancer.