INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

The nucleolus is a highly dynamic compartment of the nucleus whose size, number, and structure vary according to cell type and metabolic state. Despite this versatility, its morphological compartmentation is remarkably conserved throughout evolution: fibrillar centers (FCs), a dense fibrillar component (DFC), and a granular component (GC) have been identified in nucleoli from yeast to human, with rare exceptions. The assignment of precise functions to these morphological domains is still debated. Nevertheless, the ubiquity of the nucleolar subcompartments raises questions about the significance of the nucleolar organization. Is this spatial organization required for ribosome biogenesis to occur properly? Are the nucleolar subcompartments purely formed by the molecular mechanisms of ribosome biogenesis? Do some highly conserved nucleolar features, such as RNA polymerase I (RNA pol I) and the chromosomal context of rDNA, play a role in the organization of the nucleolus?

Some authors analyzed the nucleolar reorganization induced by drugs that inhibit particular steps of ribosome biogenesis to gain insight into the role of the transcription and maturation machineries on nucleolar formation and organization (for review, see Hadjiolov, 1985; Wachtler and Stahl, 1993). For example, treatment with drugs that specifically inhibit rDNA transcription yields nucleolar segregation, whereas chemical agents that specifically block late steps of ribosome maturation lead to nucleolar hypertrophy and invasion of granular material. Other compounds induce the mislocalization of nucleolar molecules. However, the specificity of such approaches can be questioned, because the drugs inhibiting RNA pol I transcription are likely to act on other biochemical mechanisms than rRNA synthesis (Kostura and Craig, 1986).

More recently, yeast has proved an attractive experimental system to study the molecular determinants of the nucleolar structure. The need of active RNA pol I for nucleolus formation was directly demonstrated in Schizosaccharomyces pombe thermosensitive RNA pol I mutants (Hirano et al., 1989). Saccharomyces cerevisiae strains deleted of genomic rDNA and synthesizing rRNA from plasmids were engineered, thus allowing the study of the function of the tandem organization of ribosomal genes (Chernoff et al., 1994; Nierras et al., 1997). Similarly, the role of RNA pol I in nucleologenesis was examined with mutants in which rDNA is transcribed by RNA pol II (Oakes et al., 1993, 1998). These studies concluded that ribosome biogenesis could properly occur in the absence of a normal chromosomal context. Interestingly, nucleolar regions containing the plasmid-borne rDNA as well as the rRNA-processing machinery are still detected in these strains, but their shape and location within the nucleus are different when rDNA transcription is carried out by RNA pol I or RNA pol II.

Here, we bring these studies one step further and re-explore, using S. cerevisiae mutants, the long-lasting questions of the functional compartmentation and of the dynamics of the nucleolus. We have previously shown that nucleolar subcompartments similar to those of higher eukaryotes are detected in both fission and budding yeasts (Léger-Silvestre et al., 1997, 1999). To identify determinants of nucleologenesis and nucleolar organization, we have analyzed by electron microscopy two yeast strains, L1494 and NOY 558, in which the 35S rRNA is no longer synthesized from tandemly repeated genes on the chromosome but is produced from plasmids by RNA pol I or RNA pol II, respectively. In addition to addressing the question of the role of the rDNA chromosomal context, these strains allow assessment of the influence of the transcription machinery. Electron microscopy in situ detection techniques, in conjunction with high-resolution morphology, have been chosen to precisely localize ribosomal genes, ribosomal transcripts, and proteins involved in ribosome biogenesis in these mutants. In addition, we have studied the dynamics of these nucleolar regions, as well as of wild-type nucleoli, according to rDNA transcription status. The ultrastructural data have been complemented with direct observations in living cells of a fusion protein between the nucleolar protein Gar1p and the green fluorescent protein (GFP). Our results indicate that these artificial nucleolar regions partially resemble a wild-type nucleolus, and that their assembly is directly dependent on rDNA transcription.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Media, Strains, and Culture Conditions

YEPD medium contains 1% peptone, 1% yeast extract, and 2% D-glucose. YEP-Gal is the same as YEPD, except that D-glucose is replaced by D-galactose. Synthetic medium SD contains 2% D-glucose and 0.67% yeast nitrogen base and is supplemented with the amino acids and bases required; synthetic medium SG contains 2% D-galactose instead of glucose.

The strains and plasmids used in this study are described in Table 1. The strains L1489 and L1494 were cultured in YEPG at 30°C. The strains NOY 505 and NOY 558 were cultured in YEP-Gal or in YEPD at 30°C. The temperature-sensitive strain D306-5d was cultured in YEPG at 25 or 37°C. All the cells were harvested for experiments at an A₆₀₀ between 0.4 and 0.6. 

Electron Microscopy

Freeze-Substitution Electron Microscopy. The yeasts were cryofixed with an (Ventana RMC, Tuscon, AZ) MF7200 propane jet freezer and freeze substituted at 90°C in anhydrous acetone for 2 d, followed by 5% OsO₄ in acetone at 90°C for 1 d. The temperature was then gradually increased (3°C/h). The samples were then rinsed in pure acetone, infiltrated, and embedded in epoxy resin.

Chemical Fixation. Cells were prepared according to the method of Léger-Silvestre et al. (1999) with minor modifications. They were fixed for 45 min with 4% formaldehyde with or without 0.5% glutaraldehyde in sodium cacodylate buffer (0.1 M), pH 7.2, containing 5 mM MgCl₂ at room temperature.

In Situ Detection of rDNA, rRNA, and Plasmids

The specific probe for rDNA and rRNA in S. cerevisiae was a recombinant plasmid (pBKS 35S) containing a HindIII fragment of 6.6 kb that corresponds to one 35S rDNA repeat.

The prokaryotic plasmid pBR322 was used as a specific probe to detect the yeast plasmids pRDN-wt and pNOY 103, which derive from it. Controls were made on wild-type (wt) yeasts with the pBR322 probe: no labeling was detected. The antisense external transcribed spacer 1 (ETS 1) probe (from +123 to +604) specific for the region upstream of the A₀ cleavage site (+606) of the 35S rRNA (Hughes and Ares, 1991) was used to localize the nascent transcripts. Controls were made with RNase pretreatment and with a sense ETS 1 probe: as expected, the signal was greatly decreased in both cases.

The probe plasmids pBR322 and pBKS 35S were nick translated using digoxigenin-11 (DIG-11)-dUTP (Roche Diagnostics, Mannheim, Germany) with a nick translation kit (Amersham Pharmacia Biotech, Uppsala, Sweden) according to the suppliers' instructions with a minor modification; to the recommended nucleotide mixture, we added 4.5 µl dTTP (0.375 mM in Tris, pH 8.0). The DIG-labeled antisense ETS 1 probe was synthesized by "single primer PCR." The following primer was used at 60 µM: 5'-CTTATTGAGTTTGGAAACAG-3', and 1 µg of plasmid pRDN-wt, digested by BglII, was used as template. GoldStar DNA polymerase obtained from Eurogentec (Seraing, Belgium) was used at 1 U/µg template with the following PCR conditions: 0.2 mM dATP, dCTP, and dGTP; 0.18 mM dTTP; 0.02 mM DIG-11-dUTP; 1.5 mM MgCl₂; one step at 95°C, 7 min; 20 cycles of 95°C, 45 s, 52°C, 1 min, and 72°C, 1 min; and one step at 72°C, 10 min. The DIG-labeled sense ETS 1 probe was synthesized with the same protocol using the following primer: 5'-ATGCGAAAGCAGTTGAAGAC-3', and the plasmid pRDN-wt was digested by HindIII as template.

The in situ detection of rRNA, rDNA, and plasmids was performed according to the method of Léger-Silvestre et al. (1999). To detect specifically 35S rDNA, grids were pretreated with DNase-free RNase (Roche Diagnostics) before ISH with the pBKS 35S probe. The nucleotic probes were detected with an anti-digoxigenin antibody 10-nm gold conjugate (British BioCell Research, Cardiff, UK). Controls were performed omitting the probe; no labeling was detected on these grids.

Immunocytochemistry

We used a polyclonal anti-Gar1p antibody that specifically recognized the Gar1 protein from S. cerevisiae, kindly provided by Dr. M. Caizergues-Ferrer (Laboratoire de Biologie Moléculaire Eucaryote, Toulouse, France). The monoclonal antibody directed against Ssb1p was produced by Dr. J. Broach (Princeton University, Princeton, NJ) and kindly provided by Dr. M. Nomura (University of California, Irvine, CA). The monoclonal anti-Nop1p antibody that specifically recognized S. cerevisiae Nop1 protein was a generous gift from Dr. J. Aris (University of Florida, Gainesville, FL). We also used polyclonal antibody directed against RNA pol I subunit A₁₃₅ from S. cerevisiae. This antibody was kindly provided by J. Huet (Commissari at à l'Energie Atomique, Saclay, France). To localize RNA pol II, we used commercial monoclonal antibody directed against a highly conserved heptapeptide repeat of the largest subunit of eukaryotic RNA pol II (QED Bioscience, San Diego, CA).

Polyclonal and monoclonal antibodies were revealed with gold-conjugate goat anti-rabbit and goat anti-mouse secondary antibodies, respectively (BioCell Research Laboratories).

The ultrastructural immunolocalization of the different proteins was performed according to the method of Léger-Silvestre et al. (1999). When controls were performed using secondary antibody alone, no labeling was detected on these grids.

Semiquantitative Analysis of Nop1p and Ssb1p Distribution

Micrographs were analyzed after immunolocalization of Nop1p and Ssb1p in each cell type. The number of gold particles in nucleolus and nucleoplasm was hand counted on 20-25 micrographs of independent cells taken at random. Areas of the nucleolus and nucleoplasm were estimated by cutting up the micrograph and weighting the pieces. The minimum sample size for each condition and cell component considered was determined by the progressive mean technique (confidence limit, 10%) (Williams, 1977).

In Situ Fluorescence Microscopy

To visualize the whole nucleolar structure in living cells, a fusion protein of Gar1p with GFP was used. Gar1p is an essential protein required for rRNA maturation and localizes to the dense fibrillar component throughout the nucleolus (Léger-Silvestre et al., 1999). The cDNA encoding the red-shifted GFP mutant GFP-mut2 (S65T,V68L,S72A) (Cormack et al., 1996) was amplified by PCR and subcloned in frame with the 3' end of the GAR1 coding sequence in centromeric vectors pFL38 and pFL39 (Bonneaud et al., 1991), leading to plasmids pZUT3 and pZUT4, respectively (Table 1). These plasmids also include the GAR1 promoter and transcription terminator. The fusion protein replaced wild-type Gar1p in a strain deleted of the GAR1 gene, indicating that it is fully functional. The strains NOY 505 and NOY 558 were transformed according to standard procedure (Gietz and Schiestl, 1995) with pZUT4. The other strains studied were transformed with plasmid pZUT3. The cells were cultured in SD or SG (for NOY 505 and NOY 558) in the conditions described above; 2 µl of each culture in exponential growth phase were placed on a microscope slide. The cells were observed either with a laser scanning confocal microscope (LSM 410; Zeiss, Oberkochen, Germany) or with an epifluorescence microscope (DMRB; Leica, Nussloch, Germany) equipped with CoolView camera (Photonic Science; Robertsbridge, UK).

The number of visible mininucleoli was hand counted in living NOY 558 cells transformed with pZUT4. The number of mininucleoli was recorded for 100 cells chosen at random.

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Ultrastructure of the Nucleolar Regions in Yeasts Transcribing 35S rDNA from Multicopy Plasmids

In strain L1494, the 150-200 chromosomal rDNA repeats are largely deleted and replaced by an equivalent number of plasmids bearing a single rDNA repeat, which includes the 35S rDNA transcribed by RNA pol I and the 5S rDNA transcribed by RNA pol III (Chernoff et al., 1994). The efficiency of rDNA transcription in this strain is lower than in wild-type, but 35S rRNA processing occurs normally (Nierras et al., 1997).

NOY 558 cells are unable to transcribe genomic rDNA because of the absence of Rrn7p, a protein essential for the association of RNA pol I to the promoter (Steffan et al., 1996). Synthesis of 35S rRNA is ensured by RNA pol II from a plasmid (20-50 copies per cell) bearing the transcribed region of the 35S gene under the control of the GAL7 promoter (Nogi et al., 1991). Therefore, NOY 558 cells only grow with galactose as the carbon source.

To get insight into the fine organization of the nucleolus, these strains were first prepared for electron microscopy observation by cryofixation followed by freeze-substitution, a method that has proved to yield superior preservation of nuclear ultrastructure (Léger-Silvestre et al., 1999) (Figure 1, A-C).

View larger version (153K): [in this window] [in a new window]   Figure 1.   Nuclear ultrastructure of S. cerevisiae mutants L1494 and NOY 558. Yeast strains L1489 (wild-type), L1494 (rdn, pRDN-wt), and NOY 558 (rrn7, pGAL7-RDN) were cryofixed (A-C) or chemically fixed (A'-C') and prepared for electron microscopy. The nucleolus in wild-type strain L1489 (A and A') presented the three typical components: FCs, a DFC, and a GC. In mutant L1494 (B and B'), the nucleus displayed an electron-dark region made of a fibrillar network (arrows) and a granular region (G). The nucleus of mutant NOY 558 (C and C') contained one or several electron-dark regions, referred to as mininucleoli (arrows). Two domains were identified in mininucleoli: a central compact region (arrowhead) surrounded by a diffuse region presenting mostly a fibrillar aspect. Bar, 300 nm.

As previously described in S. cerevisiae (Léger-Silvestre et al., 1999), the wild-type parental strains (L1489 and NOY 505) of both mutants displayed a crescent-shaped nucleolus of high electron density, close to the nuclear envelope and occupying one-third of the nucleus (our unpublished results for NOY 505). The three nucleolar components were recognizable: an electron-lucid zone resembling FCs, a fibrillar region of great electron density surrounding FCs and identified as the DFC, and a GC forming the rest of the nucleolus (Figure 1A for L1489).

In contrast to the parental strains, each mutant displayed a peculiar ultrastructural organization of the nucleus. In cryofixed L1494 (Figure 1B), the nucleus comprised an electron-dense structure close to the envelope, narrower and more stretched than the wild-type nucleolus. Two subregions could be clearly distinguished: a DFC (arrow), morphologically resembling the DFC of wild-type yeast, positioned at the border with the nucleoplasm, and a lighter region, which appeared mainly granular (G). No fibrillar center was detected in this structure (Figure 1B). In cryofixed NOY 558, the nucleolus was replaced by one to three electron-dense bodies (arrow) scattered in the nucleoplasm (Figure 1C), previously described as "mininucleolar bodies" or "mininucleoli" in strains equivalent to NOY 558 (Oakes et al., 1993, 1998). These substructures were made of a central compact region, which we called the "dense component" (arrowhead), surrounded by a wider region presenting a fibrillar aspect referred to as the "peripheral component." Granules dispersed between fibrils were detectable, but no granular component per se was identified (Figure 1C).

When the cells were prepared with a conventional chemical fixation more suitable for immunocytochemistry and ISH (no osmium tetroxide, hydrophilic resin), the nucleolar substructures were still recognizable, although some morphological details were lost when compared with cryofixed cells, in particular, the fibrillar organization of the mininucleoli (Figure 1, A'-C').

Ribosomal rDNA transcription sites in L1494 and NOY 558

ISH was next used to localize the ribosomal transcripts and the plasmids bearing the ribosomal genes, whereas the transcription machinery was detected using antibodies against protein A₁₃₅, a subunit of RNA pol I, or antibodies raised against wheat-germ RNA pol II.

In yeast L1494, plasmids pRDN-wt (from which 35S rDNA is transcribed by RNA pol I) were detected by ISH with the rRNA probe after RNase pretreatment. They clearly appeared clustered in several small spots within the nucleolar region (Figure 2A). RNA pol I (Figure 2B) mostly colocalized with the majority of the ribosomal transcripts (Figure 2D) in the nucleolar region, whereas RNA pol II was excluded from it (Figure 2C). To detect nascent pre-RNA, we used a probe complementary to the sequence located upstream from the early cleavage point A₀ in the 5' ETS 1 (Hughes and Ares, 1991). Equivalent probes have been successfully used to localize nascent ribosomal transcripts in other organisms (Pierron and Puvion-Dutilleul, 1996; Lazdins et al., 1997). In the wild-type parental strains L1489 and NOY505, this probe labeled the DFC throughout the nucleolus (Figure 3, A and C). L1494 cells showed ETS 1 labeling within the nucleolar region, mainly in the DFC (Figure 3B). This result indicates that in L1494, nascent ribosomal transcripts rapidly accumulate in the DFC of the nucleolar region, which may be the site of rDNA transcription.

View larger version (123K): [in this window] [in a new window]   Figure 2.   Electron microscopic localization of the plasmids, RNA pol I, RNA pol II, and 35S rRNA in L1494 and NOY 558. The plasmids bearing rDNA were detected by ISH in the nucleolar regions of both strains L1494 (rdn, pRDN-wt; A) and NOY 558 (rrn7, pGAL7-RDN; E). In NOY 558, the plasmids were precisely localized to the peripheral component. When detected by immunolabeling in L1494, RNA pol I was mainly visible in the nucleolar region (B), whereas pol II was excluded from it (C). In NOY 558, both RNA pol I and RNA pol II were localized to the nucleoplasm and the mininucleoli, the dense component excepted (F and G). In both mutants, ribosomal transcripts were concentrated in the nucleolar regions (D and H). The cytoplasmic labeling corresponds to the ribosomes. In NOY 558, the mininucleoli are designated by arrows. Bar, 300 nm.

View larger version (161K): [in this window] [in a new window]   Figure 3.   Electron microscopic detection of the ETS 1 of 35S rRNA in wild-type yeasts L1494 and NOY 558. The upstream region of the ETS 1 in 35S rRNA was localized by ISH in the nucleus of L1489, L1494, NOY 505, and NOY 558. In the parental strains (L1489 and NOY 505), the labeling was exclusively detected along the DFC (A and C). In L1494, the gold particles were positioned along the fibrillar network of the nucleolar region (B). In NOY 558, most of the gold particles were localized in the peripheral component of the mininucleoli (arrows), whereas none was detected in the dense component (arrowhead) (D). The cytoplasm was not labeled. Bar, 300 nm.

In NOY 558 nucleus, plasmids pNOY 103, in which rDNA is under the control of the GAL7 promoter, were detected with the pBR322 probe in the peripheral component of the mininucleoli (Figure 2E). RNA pol II, which transcribes 35S rDNA in this strain in addition to mRNA, was detected in the peripheral component of mininucleoli as well as in the nucleoplasm but was excluded from the dense component (Figure 2G). RNA pol I, which is transcriptionally inactive in this mutant, was dispersed throughout the entire nucleus, the dense component of the mininucleoli excepted (Figure 2F); RNA pol I location is restricted to the nucleolus in wild-type strains (Léger-Silvestre et al., 1999). In the nucleus, the ribosomal transcripts were mainly detected in the peripheral component, but some were visible in the dense component (Figure 2H). The peripheral component of the mininucleoli was labeled with the antisense ETS 1 probe (Figure 3D) but not the dense component (arrowhead). Thus, colocalization of RNA pol II, plasmid-borne rDNA, and nascent ribosomal transcripts in the peripheral component indicates that ribosomal transcription occurs in this substructure and not in the dense component.

Taken together, these results show in two different models that even though ribosomal genes are spread on separated copies of a plasmid, they assemble and are transcribed in well-defined nucleolar structures. Moreover, in both cases rDNA transcription occurs in specific subcompartments of these nucleolar structures.

Recruitment of the rRNA-processing Machinery to the Nucleolar Regions

To identify the subcompartment of the nucleolar regions in which pre-rRNA processing takes place when transcripts originate from plasmids, three proteins involved in early steps of rRNA processing, Ssb1p, Nop1p, and Gar1p, were localized by immunocytochemistry in strains L1494 and NOY 558 (Figure 4).

View larger version (169K): [in this window] [in a new window]   Figure 4.   Electron microscopic immunolocalization of Ssb1p and Nop1p in L1494 and NOY 558. In L1494, both proteins were positioned along the fibrillar network in the nucleolar region (A and B). In NOY 558, the peripheral component (arrows), but not the dense component, was labeled (C and D). Bar, 300 nm.

Ssb1p associates with snR10 and plays a putative a role in early rRNA processing (Clark et al., 1990); Nop1p is a component of the box C+D class of small nucleolar ribonucleoprotein particles (RNP), whose function is linked to pre-rRNA methylation (Tollervey et al., 1991); Gar1p is a component of the box H+ACA small nucleolar RNP (Lafontaine and Tollervey, 1999, and references therein) and is required for rRNA pseudouridylation (Bousquet-Antonelli et al., 1997).

In strain L1494, the three proteins were almost exclusively found in the nucleolar region (Figure 4, A and B; our unpublished results for Gar1p), like the ribosomal transcripts. Moreover, gold particles were often aligned along the DFC (arrows), in a manner reminiscent of the distribution of the same proteins along the DFC in wild-type nucleolus (Léger-Silvestre et al., 1999). In NOY 558, the mininucleoli contained the majority of these three proteins. The labeling was restricted to the peripheral component of mininucleoli (Figure 4, C and D; our unpublished results for Gar1p), whereas the dense component was never significantly labeled by any of the antibodies tested.

In addition, we localized Gar1p in living L1494 and NOY 558 cells with a fusion protein between Gar1p and the GFP. The chimeric protein complements deletion of the essential GAR1 gene in a manner similar to that of the wild-type gene. When introduced in the parental strain NOY 505, the fluorescent protein was exclusively observed in a compact region of the nucleus (Figure 5A'), as observed with anti-Gar1p antibodies (Girard et al., 1992). In accordance with the electron microscopy pictures, the fluorescence in L1494 cells generally extended along the nuclear envelope, forming a region that sometimes appeared as ring shaped (Figure 5B'). In NOY 558 cells, Gar1-GFP accumulated in several (one to six) small round fluorescent spots in the nucleus; most of the cells only displayed two or three spots (Figure 5C'; see Figure 10). By changing the focus, it was often possible to distinguish a dark region in the center of these spots.

View larger version (87K): [in this window] [in a new window]   Figure 5.   Detection of Gar1-GFP by laser scanning confocal microscopy in L1494 and NOY 558. In wild-type strain NOY 505, the fusion protein is specifically found in the nucleolus (A and A'). In L1494 cells, the fluorescence is restricted to a region extending along the nuclear envelope (B and B'). In NOY 558 cells grown in galactose, the fusion protein is located in several mininucleoli (C and C'). Bar, 5 µm.

These results indicate that in yeasts transcribing rDNA from plasmids, processing complexes are recruited to the site of plasmid gathering, where transcription occurs. The colocalization of constituents required for ribosome biogenesis on a fibrillar network strongly suggests that a subcompartment with the same function as the DFC in wild-type nucleolus organizes in these mutants.

Active rDNA Transcription Is Necessary for the Colocalization of the Plasmids and the Processing Complexes in NOY 558

Because the presence of the rRNA-processing machinery at the site of rDNA transcription is observed independently of the RNA polymerase involved, one may hypothesize that formation of the nucleolar regions is driven in part by association of rRNA-binding complexes to nascent rRNA. To determine whether nucleologenesis is linked to rDNA transcription, we looked at the fate of the mininucleoli in NOY 558, when the GAL7 promoter, which drives rDNA transcription from plasmids, is repressed by addition of glucose.

Arrest of 35S rDNA transcription by glucose led to a strong reduction of the peripheral component of the mininucleoli, whereas the dense component was unaffected (Figure 6A). Although the rRNA level in the nucleus dropped upon inhibition (compare Figure 7B with Figure 2H), a significant amount of ribosomal transcripts was still detected in the residual component (Figure 7B). In parallel, the ISH signal for nascent ribosomal transcripts (antisense ETS 1 probe) was dramatically lowered, as expected from inhibition of rDNA transcription (Figure 7C). The labeling was equivalent to that obtained after ISH with the sense probe (our unpublished results). Similarly to rRNA, plasmid pNOY 103 (Figure 7A) and proteins Nop1p, Gar1p, and Ssb1p (Figure 7D for Nop1p) were also detected in the residual peripheral component (arrows). But strikingly, a major portion (55-65%) of these proteins was dispersed in the nucleoplasm, in contrast to their preferential localization in the mininucleoli under permissive conditions (Figure 8). Similarly, the plasmids were also localized to the nucleoplasm (Figure 7A). The intensity of the ISH signal for the plasmids in NOY 558 under inhibition was significantly higher when compared with cells without inhibition (Figures 2E and 7E). As already discussed (Goessens, 1984; Scheer and Raska, 1987), DNA may be more accessible to the probe if not transcribed.

View larger version (84K): [in this window] [in a new window]   Figure 6.   Nuclear ultrastructure of NOY 558 according to rDNA transcription state. Inhibition of rRNA synthesis in NOY 558 (4 h in glucose) led to a strong reduction of the peripheral component (arrow) of the mininucleoli but not of the dense component (A, arrowhead). After resuming rDNA transcription by culturing the yeasts for 10 h in YEP-Gal, the mininucleoli (arrow) were identical to those observed in NOY 558 before inhibition; the dense component (arrowhead) was still visible (B).

View larger version (117K): [in this window] [in a new window]   Figure 7.   Relocation of nucleolar molecules in NOY 558 according to rDNA transcription state. After inhibition of rRNA synthesis in NOY 558 (4 h on glucose), plasmid pNOY 103 was detected both in the mininucleoli (arrow) and in the nucleoplasm (A). The nuclear labeling corresponding to rRNA was weaker than in rDNA-transcribing cells (Figure 2H) and was preferentially found in the mininucleoli (B, arrows). The upstream region of the ETS 1 of 35S rRNA was localized by ISH. The residual component (arrow) was not significantly labeled (C). Nop1p was visible in the residual component (arrow) but was also dispersed throughout the nucleoplasm (D). After resumption of rDNA transcription, plasmids (E), rRNA (F), ETS 1 (G), and Nop1p (H) were preferentially found in the peripheral component, as before inhibition (arrows). Bar, 300 nm.

View larger version (23K): [in this window] [in a new window]   Figure 8.   Semiquantitative analysis of Nop1p and Ssb1p immunolabeling in NOY 558. Distribution of gold particles present in the nucleolus (light bars) or in the nucleoplasm (dark bars) of NOY 558 (rrn7, pGAL7-RDN) after immunolocalization of Nop1p and Ssb1p is shown. The first and third bars (Gal) correspond to cells transcribing rDNA, and the second and fourth bars (Glu) correspond to cells under inhibition.

Dispersion of Gar1p in the nucleoplasm was also observed by fluorescence microscopy in NOY 558 cells expressing Gar1-GFP (Figure 9B). After 4 h on glucose, the number of mininucleoli was significantly lower, a majority of cells displaying a unique mininucleolus and <10% containing more than two (vs. 35% on galactose) (Figure 10). Diffuse staining was visible in the entire nucleus, indicating the relocation of Gar1-GFP to the nucleoplasm.

View larger version (42K): [in this window] [in a new window]   Figure 9.   Relocation of Gar1-GFP in NOY 558 and D306-5d nucleus according to rDNA transcription state. Yeast strains NOY 558 and D306-5d were transformed with a plasmid encoding Gar1-GFP. Under permissive conditions in NOY 558, Gar1-GFP was exclusively detected in several mininucleoli (A). After 4 h in glucose-containing medium, most of the cells displayed a single mininucleolus, whereas a diffuse fluorescence was detected in the whole nucleus (B). In D306-5d grown at permissive temperature, the fluorescence was restricted to the nucleolus (C). At 37°C, an intense spot was still visible, but the whole nucleus was also fluorescent (D). The dotted line indicates the boundary of the nucleus. Bar, 2 µm.

View larger version (17K): [in this window] [in a new window]   Figure 10.   Semiquantitative analysis of the number of mininucleoli in NOY 558. Distribution of the number of mininucleoli per nucleus is shown for 100 NOY 558 Gar1p-GFP cells transcribing rDNA (light bars) or after transcription inhibition (dark bars).

Ten hours after removal of glucose from the growth medium, the mininucleoli recovered their original aspect (Figure 6B), and the distribution of the nucleolar constituents was similar to that in NOY 558 without inhibition (Figure 7, E-H). The dense component remained unchanged during inhibition and resumption of rDNA transcription and was never labeled, except faintly with the rRNA probe (Figure 7F).

All these results indicate that rDNA transcription is required for the assembly of a functional nucleolar region in NOY 558.

Ribosomal DNA Transcription Is Also Required to Maintain the Structured Integrity of a Wild-Type Nucleolus

To extend the results obtained with NOY 558 to a strain transcribing 35S rDNA from the chromosome by RNA pol I, we examined the effect of rDNA transcription arrest in strain D306-5d, which bears a temperature-sensitive allele of RPA190 (rpa 190-2), an essential gene encoding the largest subunit (A₁₉₀) of RNA pol I (Wittekind et al., 1988). The rpa190-2 allele contains a UGA nonsense codon, which leads to the production of a truncated form at 37°C and to the progressive arrest of rDNA transcription.

D306-5d cells grown at 25°C displayed a nucleolus similar to that of a wild-type strain (Figure 11A). Shifting to 37°C progressively led to a dramatic nucleolar reorganization (Figure 11E). The nucleolus was segregated into three major components: a domain of low electron density, partially surrounded by high-electron-dense fibrillar regions (arrow), and a region of intermediate electron contrast. This segregation was reminiscent of the effect observed upon inhibition of RNA pol I by actinomycin D in plant and animal cells (Hadjiolov, 1985). This nucleolar remnant was disconnected from the nuclear envelope and more compact than the nucleolus of cells grown at 25°C. The ISH signal for nuclear ribosomal transcripts was dramatically reduced at 37°C in agreement with an arrest of rDNA transcription (Figure 11, compare B with F). However, rRNAs were still detected in the vicinity of the compact high-electron-dense domain of the nucleolus (Figure 11F). These RNAs did not correspond to nascent transcripts, because they were not significantly detected by the ETS 1 antisense probe (Figure 11C; labeling equivalent with the sense probe, our unpublished results). No rRNA was detected in the nucleoplasm (Figure 11G). As in NOY 558, a significant part of Nop1p, Ssb1p, and Gar1p was found in the nucleoplasm, whereas some remained in the intermediate-electron-dense domain of the segregated nucleolus (Figure 11H; see the semiquantitative analysis of Nop1p and Ssb1p in Figure 12A). This redistribution was not a direct consequence of the 37°C heat shock, because it was not observed in wild-type strain NOY 505 grown for 10 h at this temperature (Figure 12B). Accordingly, when D306-5d cells were transformed with a plasmid encoding Gar1-GFP, the fluorescence, restricted to the nucleolus at permissive temperature (Figure 9C), spread to the entire nucleus at 37°C. A single spot intensively labeled was still visible, presumably corresponding to the nucleolar remnant (Figure 9D).

View larger version (126K): [in this window] [in a new window]   Figure 11.   Electron microscopic analysis of the nucleolar dynamics in strain in D306-5d according to rDNA transcription state. At 25°C, strain D306-5d exhibited a wild-type nucleolus (A). Nuclear rRNA (B), ETS 1 (C), and Nop1p (D) were located in the nucleolus. Upon transfer to 37°C for 4 h, the nucleoli segregated into three major components (E): low-electron-dense domains (asterisk) surrounded by a high-electron-dense region (arrow) and a region of intermediate electron contrast extending separately (arrowhead). After ISH with the rRNA probe, only few gold particles could be detected at the periphery of the high-electron-dense region (F). The ETS 1 labeling was dramatically reduced (G). Nop1p was detected in the whole nucleus and accumulated in part in the low-electron-dense domain of the segregated nucleolus (H). Bar, 300 nm.

View larger version (15K): [in this window] [in a new window]   Figure 12.   Semiquantitative analysis of Nop1p and Ssb1p immunolabeling in D306-5d and NOY 505. Distribution of gold particles is shown in the nucleolus (light bars) and the nucleoplasm (dark bars) of D306-5d (rpa 190-2; A) and NOY 505 (wild-type; B) grown at 25 or 37°C for 10 h.

These results show that even when rRNA genes are in a normal chromosomal context, inhibition of ribosomal transcription leads to a dynamic spatial rearrangement of the nucleolar region and in the dispersion of the rRNA-processing machinery. rRNA synthesis appears to be one of the main requirements for nucleologenesis.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

Effects of the Absence of Tandem Organization of Ribosomal Genes on the Nucleolar Region

Ribosomal gene clustering on chromosome XII (Petes, 1979) is not necessary to form morphologically identified structures dedicated to ribosome biogenesis in S. cerevisiae, as already discussed by Oakes et al. (1993, 1998). However, as seen in L1494, in which RNA pol I transcribes plasmidic rDNA, transcription of rDNA is not sufficient by itself for organizing a wild-type nucleolus. Although location of the nucleolar region close to the nuclear envelope is conserved in these cells, this structure is not confined to one side of the nucleus as in wild-type cells, which suggests that the tandem distribution of rDNA genes on one chromosome in yeast provides a spatial constraint for the formation of the nucleolus.

Nevertheless, the ultrastructural data presented here show that the subcompartmentation of the nucleolus in three domains, namely FC, DFC, and GC, as observed in wild-type yeasts (Léger-Silvestre et al., 1999; this article), is partly conserved when rDNA transcription occurs on plasmids. Indeed, electron microscopy after cryofixation shows, in L1494 cells, a nucleolar region composed of a granular component and a fibrillar domain strongly resembling DFC in normal nucleoli (Figure 1B). Codistribution of Gar1p, Nop1p, Ssb1p, and nascent transcripts on this fibrillar network indicates that it is the place of rRNA maturation, like the DFC in wild-type yeast. Thus, the presence of a DFC in L1494 suggests that assembly of this nucleolar subcompartment is independent of rDNA organization. In contrast, neither L1494 nor NOY 558 display a fibrillar center. This observation indicates that ribosome biogenesis may occur while no fibrillar center is morphologically identified in the nucleolus. The lack of FCs in mutant yeasts containing rDNA on plasmids strongly supports the hypothesis, already discussed in higher eukaryotes, that FCs might correspond, at least in part, to the nucleolar organizer regions (NORs) (Goessens, 1984; Shaw and Jordan, 1995).

Interestingly, although NOY 558 mininucleoli are organized differently when compared with L1494 nucleolar region, the peripheral component of these substructures appears to fulfill the same functions as the DFC, because molecules involved in rRNA synthesis and maturation are concentrated in this fibrillar component. The weakness of the granular component (presumably maturing preribosomal particles in wild-type cells) may be attributed to a low rate of ribosome production in this strain, due in part to the small number of transcribed rDNA genes (20-50 plasmids per cell). Unlike wild-type nucleoli, mininucleoli also contain a central dense component to which no function could be assigned. rRNA (but not nascent transcripts containing ETS 1) was the sole nucleolar molecule that we detected in this subregion. The absence of such a substructure in L1494 demonstrates that it does not artifactually result from the distribution of ribosomal genes on plasmids and is probably linked to the transcription of ribosomal genes by RNA pol II. Accordingly, a dense core was observed in the nucleolar structure formed when RNA pol II transcribes 35S genes inserted in chromosome XII (under control of the GAL7 promoter) (Oakes et al., 1998).

Recruitment of the rRNA Maturation Complexes in Nucleolar Regions Is Consequent to rRNA Synthesis

Co-distribution of the RNA polymerase transcribing the 35S rDNA with Nop1p, Gar1p, and Ssb1p both in strains NOY 558 and L1494 shows that a tight spatial association exists between transcription and rRNA processing, whatever the RNA polymerase (I or II) engaged in transcription and independently of the chromosomal context. The reversible dispersion of the maturation machinery upon inhibition of rDNA transcription in NOY 558 provides evidence for a mechanism leading to the recruitment of the maturation components consequently to rDNA transcription. Similarly, inhibition of RNA pol I activity in mutant D306-5d (rpa190^ts) results in partial relocation of the nucleolar proteins to the entire nucleus. These data are in agreement with investigations on higher eukaryotes showing that active NORs are required to target nucleolar nonribosomal proteins to a defined position, where a functional nucleolus appears (Hernandez-Verdun et al., 1991). Prenucleolar bodies occurring during the first step of nucleologenesis at the end of mitosis behave similarly: the sole presence of NORs is not sufficient for their targeting to occur; active rRNA synthesis is required (Benavente et al., 1987; Azum-Gelade et al., 1994).

It is noteworthy that rDNA transcription by RNA pol II is able to recruit the maturation machinery normally associated with RNA pol I. This result indicates that the specific maturation pathway for rRNAs is not necessarily coupled to transcription by RNA pol I (Sisodia et al., 1987; Lo et al., 1998). The situation is different when analyzing the specificity of mRNA maturation: indeed, pre-mRNA synthesis by RNA pol I or RNA pol III leads to unprocessed transcripts suggesting that accurate maturation of mRNAs is linked to transcription by RNA pol II (Lo et al., 1998). A₁₃₅ distribution throughout the entire nucleus in NOY 558 indicates that RNA pol I accumulation is not required at sites of nucleolar emergence, suggesting that RNA pol I has no accessory role in the recruitment of the nucleolar proteins or in rRNA maturation.

The accurate recruitment of the processing machinery, when RNA pol II transcribes rDNA, strongly suggests that the ribosomal transcripts, rather than RNA pol I, are necessary in this process. Accordingly, it was shown that the RNA-binding domain of Gar1p was necessary and sufficient for nucleolar targeting of this protein (Girard et al., 1994). Also, Verheggen et al. (1998) have recently reported that during the de novo nucleolar building in Xenopus laevis embryos, the presence of pre-rRNAs of maternal origin rather than the onset of rDNA transcription is critical to organize the nucleolar domain.

Some Nucleolar Components Persist within Nucleolar Remnants under Inhibition of rDNA Transcription

Upon inhibition of rDNA transcription, nucleolar remnants containing both partially matured rRNA transcripts and nucleolar proteins were identified in NOY 558 and D306-5d. The absence of ETS 1 labeling in the residual peripheral component in NOY 558 at this stage indicates that the detection of rRNA does not reflect leakage of the GAL7 promoter (Adams, 1972).

Similarly, fibrillar nucleolar remnants have been observed in Physarum polycephalum during mitosis, a phase at which rDNA transcription is arrested. These structures contain rDNA, specific nucleolar proteins, as well as rRNA without the ETS 1 sequence and participate in nucleolus reconstruction after mitosis (Pierron and Puvion-Dutilleul, 1996). Similarly, in NOY 558 and D306-5d under inhibition, rRNA intermediates synthesized before inhibition may cluster in a specific nuclear region where they stay associated with part of the processing complexes. Whether such organizing regions correspond to specific nuclear compartments and what is responsible for rRNA coalescence remain to be established.

Coalescence of the Plasmids in the Nucleolar Region

Although rRNA synthesis might provide a template for the rRNA processing machinery and therefore trigger the assembly of the nucleolar structure at the transcription sites, it remains to be understood how the rDNA-bearing plasmids coalesce to form a limited number of structures in the nucleus. The dynamic redistribution of plasmids and the reduction in the number of mininucleoli observed upon inhibition of rDNA transcription clearly indicate that, in NOY 558, plasmid clustering depends on active transcription of the ribosomal genes by RNA pol II. One may hypothesize that mininucleoli in NOY 558 result from the concentration of molecules involved in rRNA synthesis (plasmids and maturation machinery) in transcription "factories", as described for mRNA in mammalian cells. However, in yeasts, RNA pol II has not been found to be located in foci (Elliott et al., 1992; Léger-Silvestre et al., 1999).

In contrast to NOY 558, the nucleolar region in L1494 is in close contact with the nuclear envelope, as the nucleolus in wild-type cells. This difference might be linked to rDNA sequences, present in pRDN-wt (L1494) and not in pNOY 103 (NOY 558), that could interact with nuclear substructures (Planta, 1997): untranscribed regions of rDNA and the 5S gene. However, the presence of the 5S gene on a plasmid in which rDNA is transcribed by RNA pol II is not sufficient to influence the organization of the nucleolar regions; such mutants display mininucleoli similar to those observed in NOY 558 (Oakes et al., 1998). In addition, the position of the ribosomal genes in a normal chromosomal context in strain D306-5d does not prevent a strong change in the position of the nucleolus upon inhibition of RNA pol I. This result suggests that rDNA cis elements alone are not directly bound to structures near the nuclear envelope, and that active transcription is required for the peripheral location of the nucleolus. Some of the transcription factors, which form a complex with RNA pol I on the rDNA promoter during transcription, may ensure interactions between rDNA and the vicinity of the nuclear envelope.

Nucleolar Segregation Induced by Inhibition of rDNA Transcription

For the first time, nucleolar segregation is described in S. cerevisiae. In mammals, segregation of nucleolar components takes place several hours after administration of drugs such as antibiotics and intercalating agents (for review, see Hadjiolov, 1985). All these inhibitors, like actinomycin D, block RNA synthesis by altering the structure of the DNA template. Low doses of actinomycin D preferentially inhibit rRNA synthesis and cause rapid detachment of growing pre-RNP fibrils from the rDNA axis, as shown by chromosome-spreading studies (Puvion-Dutilleul and Bachellerie, 1979). In yeast D306-5d, inhibition originates from a progressive limitation of the pool of intact RNA pol I complexes. It is noteworthy that the morphological consequences of rDNA transcription inhibition in D306-5d are similar to those observed in higher eukaryotes using actinomycin D. This result strongly supports the general consensus that nucleolar segregation reflects the arrest of transcription activity (accompanied by two others events: retained capacity of processing primary preribosomes and delayed release of nascent ribosomes from the nucleolus; for review, see Hadjiolov, 1985) rather than the action of drugs on rDNA.

Conclusion

This set of ultrastructural data shows that nucleolar subcompartmentation is partly conserved in S. cerevisiae mutants in which rDNA is transcribed from plasmids. From the data presented in this paper, we hypothesize that the formation and spatial organization of the nucleolus in S. cerevisiae result from the recruitment of the transcription and maturation machineries around the single NOR localized on chromosome XII. This clustering clearly depends on the state of activity of the ribosomal genes: ribosomal transcripts represent a primary determinant for nucleologenesis.