RNA polymerase I–specific subunits promote polymerase clustering to enhance the rRNA gene transcription cycle

In all living organisms, multisubunit DNA-dependent RNA polymerases are built around a common set of five core subunits (α₂ββ’ω in eubacteria; Minakhin et al., 2001) In archeal and eukaryotic enzymes, five additional subunits (related through common ancestry) are required, leading to a 10-subunit extended core (Cramer et al., 2008; Kwapisz et al., 2008). In eukaryotes, three nuclear RNA polymerases are found, each transcribing a dedicated set of genes. RNA polymerase II (Pol II) transcribes all protein-coding genes and most of the small nuclear RNA genes. RNA polymerase III (Pol III) is transcribing 5S ribosomal RNA (rRNA) genes, tRNA genes, and some noncoding RNA genes. RNA polymerase I (Pol I) produces a single transcript, the large polycistronic precursor (35S pre-rRNA in yeast) processed in multiple successive steps into the mature rRNAs (in yeast, 25S, 18S, and 5.8S rRNAs). Pol II is composed of 12 subunits, including a 10-subunit extended core and a peripheral heterodimer of subunits Rpb4/Rpb7. Rpb4 and Rpb7 have counterparts in Pol I and Pol III, Rpa43/Rpa14 and Rpc17/Rpc25, respectively. Pol I and Pol III have counterparts to each of the 12 Pol II subunits and, in addition, two and five specific subunits, respectively, which may be responsible for some of the specific functions of these two enzymes. In this study, we focus on Pol I–specific subunits Rpa34 and Rpa49, which have structural homology with the Rap74/Rap30 subunits of the Pol II transcription factor TFIIF (Kuhn et al., 2007).

Pol I is the most active and abundant RNA polymerase in eukaryotes. Its enormous transcriptional output can best be visualized using the DNA spread method previously developed by Miller et al. (1969), where the 35S rRNA genes (rDNA) adopt a “Christmas tree” conformation. In Saccharomyces cerevisiae, the rDNA is made of a 6.9-kb-long gene repeated ∼150 times, of which about half is transcribed. Transcription of the rDNA is one to two orders of magnitude greater than the transcription of the rest of the genome. In exponentially growing cells, 0.01–10 Pol II enzymes are found per open reading frame (Bon et al., 2006), whereas >120 Pol I molecules can be detected (by Miller spread; Miller et al., 1969) at the transcribed region of each active rDNA repeat (Osheim et al., 2009).

Pol I activity is associated with the largest nuclear body, the nucleolus, where all earlier steps of ribosome biogenesis take place. Assembly of the nucleolus is proposed to be a self-organized process, initiated by production of rRNA (Trumtel et al., 2000; Misteli, 2001; Hernandez-Verdun et al., 2002). In budding yeast, Pol I is essential, but very elegant genetic analyses have been used to show that Pol II–dependent 35S rRNA expression, albeit very inefficiently, can rescue Pol I minus mutants (Nogi et al., 1991). Importantly, in mutant cells where Pol I is inactivated and Pol II is artificially producing rRNA, nucleolar structures are still detectable, but with massive alterations (Oakes et al., 1998; Trumtel et al., 2000). Therefore, rRNA synthesis achieved from its dedicated transcriptional apparatus is important for formation of the nucleolar structure.

In this study, we focus on the role of two Pol I–specific subunits, Rpa49 and Rpa34, in Pol I transcription and nucleolar morphology. Pol I lacking Rpa34/Rpa49 has been shown to be active in transcription but impaired in elongation (Huet et al., 1975). In vivo, rpa49 and rpa34 deletion mutants are viable, but are defective in 35S rRNA production (Liljelund et al., 1992; Gadal et al., 1997). Structural insight and biochemical assays have suggested that Rpa34/Rpa49 form a heterodimer behaving like built-in elongation factors (Kuhn et al., 2007). Another clue to Rpa49 function came from the observation that in rpa49 deletion mutants, Rrn3 (TIF-IA in mouse), an essential Pol I transcription factor, is recruited to the promoter less efficiently and fails to dissociate from elongating polymerase (Beckouet et al., 2008). In this study, we demonstrate that the Rpa34/Rpa49 heterodimer is functionally conserved in evolution and required for nucleolar assembly. Using Miller spreads to analyze Pol I distribution in the absence of Rpa49 on single transcribed genes, we show that there is a decreased clustering of elongating Pol I complexes. 3D modeling of two neighboring Pol I complexes unveil how such association can regulate rRNA gene transcription cycle.

We investigated the functional conservation of specific Pol I subunits from budding yeast to human (Fig. 1 A). In budding yeast, Rpa49 and Rpa34 are not essential for viability, and their amino acid sequences are poorly conserved through evolution. Orthologues of Rpa49 and Rpa34, PAF53 and PAF49 in mouse (Hanada et al., 1996; Yamamoto et al., 2004) and hPAF53 and CAST in human (also called ASE-1 or hPAF49; Panov et al., 2006), respectively, have a conserved domain organization with S. cerevisiae subunits but low sequence similarity. In budding yeast and mouse, Rpa49/Rpa34 orthologues are easily dissociated from the enzyme by high salt treatment (Huet et al., 1975; Hanada et al., 1996), and as a result, these two polypeptides have been classified either as Pol I–associated factors in mouse or as Pol I subunits in S. cerevisiae and human. A previous complementation attempt has established that Schizosaccharomyces pombe Pol I subunit Rpa51, hereafter called Sp-Rpa49, but not mouse PAF53, can complement rpa49Δ cold sensitivity (Nakagawa et al., 2003). Because Rpa49 and Rpa34 function as a heterodimer (Kuhn et al., 2007; Beckouet et al., 2008), we tested the coexpression of human CAST and hPAF53 in a complementation assay. We first investigated the localization of human CAST when expressed in wild-type (WT) budding yeast, with or without hPAF53 expression. CAST is nuclear when expressed alone but predominantly localized in the nucleolus when coexpressed with hPAF53 (Fig. 1 B). Starting from a yeast strain bearing a TAP-tagged Rpa190 (Rigaut et al., 1999), Pol I largest subunit, we purified Pol I to test the association of CAST with budding yeast Pol I. CAST is coimmunoprecipitated with budding yeast Pol I only when coexpressed with hPAF53 (Fig. 1 C). These results suggest that hPAF53/CAST heterodimer is replacing the Rpa49/Rpa34 heterodimer in budding yeast Pol I. We then investigated hPAF53 localization with or without CAST coexpression. No YFP-hPAF53 fusion protein was detected when it was expressed alone, but it was present in the nucleolus when coexpressed with CAST (Fig. 1 D). To test the idea that CAST stabilizes hPAF53 when coexpressed in budding yeast, we quantified hPAF53 by immunodetection in whole cell extracts with or without CAST (Fig. 1 E). hPAF53 is barely detectable in budding yeast extracts, even if expressed from the strong PGK1 promoter, but it accumulates when coexpressed with CAST. Interestingly, CAST depletion in mouse and human cells leads to concomitant disappearance of PAF53, whereas the level of other Pol I subunits is unaffected (unpublished data). Therefore, our results suggest that coexpression of CAST and hPAF53 is required for hPAF53 stabilization.

We next tested the ability of the human subunits to complement the rpa49Δ and/or rpa34Δ phenotypes by plasmid-shuffling assays (see Materials and methods). hPAF53 on its own is unstable; therefore, we expressed hPAF53 under the control of the strong PGK1 or the very strong TEF1 promoter. rpa49Δ cold sensitivity was corrected by hPAF53 only when the pTEF1 promoter was used (Fig. 1 F). rpa34Δ has little growth defect, but it is hypersensitive to caffeine (1,3,7-trimethylxanthine; Beckouet et al., 2008). We assayed the ability of CAST, hPAF53, or both to suppress caffeine sensitivity. CAST and hPAF53 complemented the rpa34Δ mutant phenotype only when coexpressed (Fig. 1 G). Therefore, our results suggest that the CAST/hPAF53 heterodimer can substitute for the endogenous budding yeast Rpa49/Rpa34 within Pol I and can perform the function of the S. cerevisiae heterodimer in vivo.

Fission yeast Sp-Rpa49 can substitute for budding yeast Rpa49 (Nakagawa et al., 2003). A recent database search revealed that the S. pombe SPBC11G11.05 open reading frame (hereafter named Sp-Rpa34) has weak but significant sequence homology with Rpa34 (Beckouet et al., 2008). We cloned and expressed Sp-Rpa34 in budding yeast and observed its nucleolar localization only when coexpressed with Sp-Rpa49 (Fig. 2 A). Because of a slight toxicity of Sp-Rpa34 for S. cerevisiae, its effect on caffeine sensitivity of rpa34Δ was unclear. We next tested complementation by Sp-Rpa34/Sp-Rpa49 of the rpa34Δ mutant in an rpa135-D395N double-mutant background in which Rpa34 is essential for growth (Beckouet et al., 2008). When expressed alone, Sp-Rpa34 was unable to complement the Rpa34 essential function, but the coexpressed Sp-Rpa34/Sp-Rpa49 heterodimer restores growth (Fig. 2 B). Our results suggest that Sp-Rpa34 is the fission yeast counterpart of budding yeast Rpa34. We next deleted Sp-Rpa34 in S. pombe. The Sp-rpa34Δ mutant fission yeast was viable with only a slight growth defect (Fig. 2 C), similar to the rpa34Δ mutant in S. cerevisiae. Our previous data demonstrated that double-deletion mutants rpa34Δ and rpa49Δ are phenotypically equivalent to the rpa49 single-deletion mutant and that Rpa49 overexpression can suppress the defects associated with rpa34Δ (Beckouet et al., 2008). The Sp-rpa49Δ single-deletion mutant and the double mutant Sp-rpa49Δ Sp-rpa34Δ grow similarly (Fig. 2 D). Sp-rpa34Δ has a strong sensitivity to mycophenolate and 6-azauracil compared with WT fission yeast. Both drugs deplete the pool of NTP in vivo and have a strong effect on strains with defects in RNA polymerase elongation. Sensitivity to both drugs is suppressed by overexpression of Sp-Rpa49 (Fig. 2 E). Therefore, our results strongly suggest that in S. pombe, both Pol I–specific subunits are conserved and functionally exchangeable with the budding yeast Rpa49/Rpa34 heterodimer.

We showed that Rpa49 overexpression could suppress all known phenotypes of rpa34Δ, suggesting that Rpa34 function is restricted to stabilization of Rpa49 (Beckouet et al., 2008), but in higher eukaryotes, the orthologue of Rpa34 has several specific functions of its own (Yamazaki et al., 1999; Yamamoto et al., 2004; Panov et al., 2006). Our sequence analysis reveals that these polypeptides share a lysine-rich C-terminal domain. This domain is not required for interaction with Rpa49 (Beckouet et al., 2008), and we hypothesized that this feature could be responsible for a function unrelated to Rpa49. To verify this hypothesis, we decided to perform a search for all possible genetic backgrounds in which rpa34Δ is lethal. We performed a global genetic interaction mapping (GIM; Decourty et al., 2008) using rpa34Δ as bait. We confirmed two previously identified synthetic lethal interactions, top1Δ and rpa14Δ (Gadal et al., 1997), and identified two novel genetic backgrounds, gcr2Δ and stb5Δ, in which rpa34 became essential (Fig. 3 A). However, Rpa49 overexpression could suppress the rpa34Δ phenotype in all tested backgrounds. Moreover, the truncated version of Rpa34 lacking the C-terminal domain was functional, revealing no growth phenotype associated with the absence of this particular domain.

We next tested whether the nucleolar localization of Rpa34 also depends on Rpa49 in budding yeast. In contrast to CAST and Sp-Rpa34, Rpa34 remains in the nucleolus even in the absence of Rpa49 (Fig. 3 B). We next tried to identify which domain of budding yeast Rpa34 could behave as a nucleolar localization signal functioning independently from Rpa49. The nucleolar localization of Rpa34 in Rpa49Δ background is lost upon truncation of the C-terminal domain (Fig. 3 C). The nucleolar targeting of Rpa34 C domain in S. cerevisiae appears to be species specific. Despite the apparent similarity between the C-terminal domains of Rpa34, CAST, and Sp-Rpa34, the nucleolar targeting of CAST and Sp-Rpa34 are strictly dependent on the coexpression of their respective binding partner, PAF53 or Sp-Rpa49. In contrast, Rpa34 in S. cerevisiae has two nucleolar targeting mechanisms. One depends on Rpa49 association, the other on the Rpa34 C-terminal domain.

We have shown that the Pol I–specific Rpa49/Rpa34 heterodimer is functionally conserved from yeast to humans. We hypothesized that Pol I–specific subunits might be responsible for the two most prominent features of Pol I transcription: nucleolar assembly and high RNA polymerase loading rate per gene. We first focused on nucleolar alterations caused by the absence of Pol I–specific subunits. We labeled the nuclear periphery using GFP-Nup49, a nuclear pore protein, and the nucleolus using mCherry-Nop1, an abundant nucleolar protein, in WT, rpa34Δ, and rpa49Δ cells (Fig. 4 A). Nucleolar alterations in rpa34Δ and rpa49Δ cells were determined by thresholding the fluorescent nucleolar signal detected using confocal microscopy in a large number of cells (Berger et al., 2008). Because the cumulative distribution function accurately describes volume distribution in large samples, we used this representation to evaluate the nucleolar size for each mutant (Fig. 4 B). The nucleolar size increases considerably in rpa49Δ cells, suggesting that the integrity of nucleolar structure is compromised (Fig. 4 A). We next performed an ultrastructural study of the nucleus in rpa49Δ and rpa34Δ mutants. For optimal preservation of the nuclear structure, chemical fixation was avoided, and the cells were cryoimmobilized before low temperature substitution. The nucleolar ultrastructure is altered in both mutants (Fig. 4, C and D). In rpa34Δ cells, the nucleolus is decondensed with areas of low electron density. In rpa49Δ cells, the nucleolar morphology is dramatically altered: its characteristic crescent shape disappeared, and the dense material redistributed in the full volume of the nucleus.

Cryofixation also allows the direct observation of individualized cytoplasmic ribosomes. The number of ribosomes in rpa34Δ cells is similar with WT, but it is significantly reduced in rpa49Δ cells (Fig. 4 E). To investigate a defect in translation, we performed a polysome profile in WT and the rpa49Δ mutant. Interestingly, we detected polysomes in both backgrounds, but with the appearance of half-mers in rpa49Δ cells, reminiscent of a 60S biogenesis defect (Fig. 4 F). Such a defect was previously observed in a Pol I mutant strain defective for rRNA elongation (Schneider et al., 2007). Therefore, the morphological analysis demonstrates that Rpa49 is required for normal nucleolar morphology and efficient ribosome biogenesis. The rpa49Δ strain contains fewer 60S subunits relative to 40S, which is reminiscent of a Pol I elongation defect.

In the rpa49Δ mutant, rRNA production still sustains growth, but the nucleolus appears fully decondensed. To spatially constrain rRNA production, we decided to use mutant cells with a decreased number of rDNA repeats (Cioci et al., 2003). rDNA copy number can be stabilized at 25 copies by Fob1 deletion without significant growth defect (Cioci et al., 2003). To support normal rRNA production from such a low number of rDNA genes, the number of Pol I complexes per gene is significantly increased (French et al., 2003). We first demonstrated that double-mutant rpa49Δ fob1Δ strains are viable at 30°C, having either 190 or 25 rDNA copies (Fig. 5 B). To confirm this surprising result, we establish that rDNA copy number in both strains is unaffected by rpa49 deletion (Fig. 5 A). Growth of the rpa49Δ strain is severely impaired at 25°C (Liljelund et al., 1992). Comparing growth of the rpa49Δ fob1Δ strains containing 25 versus 190 rDNA copies, we noted that the cold sensitivity is alleviated upon reduction in the number of rDNA repeats (Fig. 5 B).

Quantitative fluorescence analysis revealed that in presence of Rpa49, the reduction from 190 to 25 rDNA copies resulted in a nucleolus of normal size (Fig. 5 C). Electron microscopy observations showed that the nucleolus in the 25 rDNA copies background is characterized by an identifiable crescent-shaped area of fibrillo-granular material adjacent to the nuclear envelope (Fig. 5 D). Interestingly, in the Rpa49 deletion strain with 25 rDNA copies, but not with 190 copies, a crescent-shaped nucleolus is detectable in most cells (60%), as evaluated by fluorescence microscopy (Fig. 5 C). Similarly, the nucleolar morphology is partially restored at the ultrastructural level because a dense structure occupying one third of the nuclear surface and in contact with the nuclear envelope can be detected (Fig. 5 D). These data suggest that a decrease in the rDNA copy number is enough to establish a partial nucleolar organization in the absence of Rpa49.

Having shown that nucleolar assembly is dependent on Rpa49, we next evaluated whether a high RNA Pol I loading rate per gene requires Rpa49. Miller spreads are the only technique allowing counting of individual Pol I on a single gene (Miller et al., 1969). However, the ability to quantify and accurately localize polymerase on a single gene relies on obtaining high contrast chromatin spreads from mutant yeast cells. The chromatin spreading technique was set up in our laboratory using strain NOY886, which contains 42 actively transcribed rDNA copies (Fig. 6 A; French et al., 2003). Because of the very high loading rate in mutant backgrounds with reduced rDNA copy number, polymerases often appear as a continuous track, making accurate counting of single polymerases difficult. We took advantage of negatively stained Miller spreads obtained with low frequency (see Materials and methods) to demonstrate that continuous tracks are best analyzed by assuming the presence of a polymerase every 12 nm along such track (Fig. 6 C). We next analyzed Pol I density and distribution on the rRNA genes in 25 or 190 rDNA copy strains with or without Rpa49 using Miller spreads (Fig. 7 A). In the presence of Rpa49, we observed 62 ± 11 Pol I complexes per single rRNA gene in the 190 rDNA copy strain. However, in this strain, we were unable to identify rDNA in chromatin spreads in the absence of Rpa49, probably because of a very low Pol I occupancy of the rDNA. We next analyzed chromatin spreads in strains with 25 rDNA copies and identified single transcription units both in the presence and absence of Rpa49 (Fig. 7 A). We detected 122 ± 17 Pol I complexes per gene in the presence of Rpa49 and 36 ± 5 Pol I complexes per gene in its absence. We then measured the distance between adjacent polymerases to evaluate whether the Pol I molecules were randomly distributed on the rRNA genes (Fig. 7 B). The distances between polymerases depend intrinsically on the Pol I density on each rRNA gene, which varies between strains. Therefore, for each polymerase load, we first compared the empirical measurements of distance between polymerases with the corresponding random distribution predicted (Fig. 7 B). In all cases, the observed distance between polymerases was shorter than that predicted for the random distribution. These results suggest clustering of polymerases. We have introduced a new parameter in our simulations to take this effect into account. When 60% of polymerases are modeled as being in direct contact with each other and 40% are randomly distributed, the simulated data are in good agreement with our experimental measurements in the case of the 190 and 25 rDNA copy strains in the presence of Rpa49 (Fig. 7 C). However, in the 25 rDNA copy/rpa49Δ strain, we measured larger distances between polymerases compared with the predictions of the model. Therefore, we propose that the interactions between successive Pol I complexes are strongly reduced in the absence of Rpa49, accounting for the reduction in the number of adjacent Pol I complexes. Therefore, Rpa49 is required to ensure high density of Pol I per rDNA gene. Furthermore, our data suggest that Rpa49 is involved in establishing contact between adjacent Pol I complexes.

Rpa49 seems to be required for interaction between adjacent transcribing polymerases. To confirm the requirement on Rpa49 for contact between adjacent polymerases, we coexpressed in a single diploid cell two of Pol I complex’s largest subunits tagged with different epitopes, Myc and HA. By purifying HA-tagged Pol I from the chromatin-bound fraction, we could evaluate the amount of polymerase dimers or multimers by assessing the extent of copurification of Myc-tagged Pol I. In WT cells, after formaldehyde cross-linking and with extensive DNase treatment, we could detect frequent Pol I dimers (Fig. 8 A). In absence of Rpa49, the proportion of Pol I dimers or multimers is dramatically reduced (Fig. 8 A, compare lane 5 with lane 6).

Using Miller spreads and coimmunoprecipitations, we show that elongating Pol I enzymes are in frequent contact in an Rpa49-dependent manner. To characterize potential molecular interactions occurring between successive polymerases, we modeled the structure of a Pol I dimer in an elongation state on DNA (Fig. 8 D). We used a cryo–electron microscopy density map of Pol I obtained as described previously (De Carlo et al., 2003) onto which the Rpa43 and Rpa49/Rpa34 heterodimer are localized far away from each other in different parts of the Pol I complex (Fig. 8 B; Bischler et al., 2002; Kuhn et al., 2007). To model the DNA upstream and downstream of the Pol I transcription bubble, we used the atomic structure of the corresponding RNA Pol II elongation complex (Protein Data Bank accession no. 2YU9; Fig. 8 C; Wang et al., 2006). A second elongation complex was then positioned according to the following constraints: (a) enzymes were placed in contact, (b) the intervening DNA connecting the two polymerases was set as linear, and (c) this linker DNA was modeled as B-DNA to find the rotational register between the two elongating polymerases (Fig. 8 D). The minimal distance between two successive active sites was found to be of 34 bp, which is in good agreement with ExoIII footprinting experiments on RNA Pol II elongation complexes (Samkurashvili and Luse, 1998). Interestingly, in this model, the Rpa49/Rpa34 subunits of one polymerase are brought in close proximity to the Rpa43/Rpa14 heterodimer of the following Pol I. Genetic evidence previously suggested that Rpa49 and Rpa43 are functionally linked (Beckouet et al., 2008). To test whether Rpa49 directly associates with Rpa43, we coexpressed GST-tagged Rpa49 and HA-tagged Rpa43 in Escherichia coli. Because Rpa49 and Rpa34 form heterodimers, we used GST-tagged Rpa49 and HA-tagged Rpa34 as positive controls. Upon purification on glutathione Sepharose beads, both Rpa34 and Rpa43 copurify with GST-Rpa49 (Fig. 8 E, lanes 2 and 6). In contrast, HA-tagged Rpa43 or HA-tagged Rpa34 are not copurified with GST alone (Fig. 8 E, lane 4 and 8). Therefore, genetic and biochemical assays suggest a direct interaction between Rpa43 and Rpa49.

In this study, we addressed the function of two Pol I-specific subunits, Rpa49 and Rpa34. We showed that heterodimers formed by human or fission yeast orthologues are functional in budding yeast, demonstrating a functional conservation of these two Pol I–specific subunits in the course of evolution. We also showed that these subunits are playing an important role in the formation of the nucleolus and are required to produce a high density of elongating Pol I on rDNA genes.

All eukaryotic cells exploit three distinct RNA polymerases to transcribe their nuclear genome. Of these, Pol I is the most divergent enzyme, probably because of its highly specialized function: it transcribes only a single gene encoding large rRNAs. Our data demonstrate that despite considerable variation in sequence, two of the Pol I–specific subunits forming a heterodimer, Rpa49/Rpa34, have orthologues throughout the eukaryotic domain, suggesting that functions of this dimer are evolutionally conserved. Our results show that overexpression of Rpa49 can rescue the growth defects in genetic backgrounds in which rpa34Δ is lethal. The C-terminal domain of Rpa34 is responsible for nucleolar localization, and it functions independently from Rpa49. Nucleolar localization sequences are known to interact with various nucleolar factors (Lechertier et al., 2007), suggesting the possibility of a direct interaction between the C-terminal domain of Rpa34 and S. cerevisiae nucleolar proteins. Indeed, several potential partners of Rpa34 have been identified in an unbiased two-hybrid interaction screen between Rpa34 and the complete genome. In this screen, three nucleolar proteins, Gno1, a cofactor of the RNA helicase Prp43, and two components of the C/D snoRNP complexes Nop56 and Nop58 were shown to interact with Rpa34 (Beckouet et al., 2008). Therefore, the C-terminal domain is involved in the recruitment of species-specific binding partners, which are responsible for nucleolar targeting.

The nucleolus is the largest nuclear domain, the integrity of which depends directly on rRNA production (Trumtel et al., 2000; Hernandez-Verdun et al., 2002). This structure is conserved in all eukaryotes, and most nuclear steps of ribosome biogenesis take place within the nucleolus. In the absence of Rpa34 and especially Rpa49, nucleolar assembly is severely compromised. Importantly, the altered nucleolar morphology in rpa49-null strains correlates with a significant reduction in the number of ribosomes per cell. When we artificially decreased the number of rDNA copies from 190 to 25, in the absence of Rpa49, we observed a partial nucleolar reassembly. Therefore, we propose that the initial event required for self-organization of the nucleolus is a spatially constrained rRNA synthesis, which initiates the chain of events leading to formation of nucleolar structures.

A previous study revealed a close relationship between Rrn3, the two heterodimers Rpa49/Rpa34, and Rpa43/Rpa14 (Beckouet et al., 2008). The release of the Pol I–specific initiation factor Rrn3 during elongation requires Rpa49, despite diametrically opposite locations within a single Pol I complex (Beckouet et al., 2008). While our work was in progress, an elegant structural study established that the Rpa49 and Rpa34 heterodimer binds DNA and stimulates polymerase-intrinsic RNA cleavage (Geiger et al., 2010). We now show that Rpa49 is required to allow contact between adjacent polymerases. 3D modeling of adjacent contiguous polymerases gives a hint as to how Rrn3 could be released. Rrn3 and its interacting partner Rpa43 are in direct contact with Rpa49/Rpa34 of the neighboring enzyme. Indeed, we know that mutations of Rpa43 behave as extragenic suppressors of the rpa49Δ phenotype, and Rpa49 and Rpa43 when coexpressed in E. coli are interacting, arguing for a physical contact between both subunits (Beckouet et al., 2008).

From our model, we propose that adjacent polymerases form a head to tail “camel caravan” in which Rpa49 from one Pol I molecule is in direct contact with Rpa43 from the neighboring molecule. These interactions determine the distance between successive Pol I complexes. In the absence of Rpa49, adjacent complexes are unable to interact, and this could lead to the lack of Rrn3 release (Beckouet et al., 2008) and to the increased distance between polymerases (this study). All of these data together strongly suggest that the two Pol I–specific subunits Rpa49 and Rpa34 are essential for nucleolar assembly and formation of “Pol I caravans,” two unique Pol I features. We now need to understand how contacts between adjacent Pol I, promoted by Rpa49, could affect yeast nucleolar integrity, impacting on rDNA organization in the nuclear space.

Yeast strains used in this study are listed in Table S1, and oligonucleotides are listed in Table S2. Yeast strains were constructed by meiotic crossing and plasmid DNA transformation. S. cerevisiae–null alleles with KanMX4 insertions were obtained from EUROSCARF and were checked using ad hoc PCR amplifications. The Sp-rpa34Δ gene deletion in S. pombe strain BENSP1-1a was performed by homologous recombination using PCR-amplified fragments with oligos 746 and 747 with plasmid p29802 as template. Sp-Rpa49Δ (rpa51Δ) was obtained as previously described using plasmid pYN1003usds (Nakagawa et al., 2003). pDONR201 (Invitrogen), pREP41 (Basi et al., 1993), pRS413, pRS423, pRS425 (Sikorski and Hieter, 1989), pVV200, pVV208 (Van Mullem et al., 2003), p29802, pUN100-mCherry-NOP1 (mCherry-NOP1) and pASZ11-CFP-NUP49 (CFP-NUP49; Berger et al., 2008), pFL36-CII, pGID-CII (Berger et al., 2007), pOG5-RPA34 (Gadal et al., 1997), pENTR-PAF53, pENTR-CAST, pENTR-HA-CAST (Panov et al., 2006), pOG1-A34 (Beckouet et al., 2008), pENTR-RPA49 (Beckouet et al., 2008), pYN1003usds (Nakagawa et al., 2003), and pNOY373 (Wai et al., 2000) were previously described. pEB5 and pENT-Rpa34DC58 were provided by E. Bertrand and F. Beckouët, respectively.

pRS425-P_PGK was constructed by ligating the NdeI fragment from pVV200 in the SmaI site of pRS425. pCNOD08, a gateway vector that allows strong transcription from the TEF1 promoter, was constructed in three steps. First, the 1,281-bp XhoI–NcoI fragment of pRS425-P_PGK containing promoter P_PGK and attR1 lambda recombination site was inserted into XhoI and NcoI–digested vector pFSKB3X (GTP technology) to generate plasmid pCNOD02. Second, promoter P_TEF1 was amplified by PCR using oligonucleotides 439 and 440 and yeast ODN1-1a genomic DNA as template. The XbaI–XhoI-digested PCR product was cloned into the same site of pCNOD02 to generate plasmid pCNOD05. The 1,281-bp NcoI–XhoI fragment of pCNOD05 was inserted into pRS425-pPGK digested with NcoI and XhoI generating pCNOD08. pRS423-P_PGK was constructed by subcloning the DraIII–SacI fragment from pRS425-P_PGK in the same site of pRS423. pCFP-Nup49-His was constructed by ligating the fragment from HindIII (filled in with Klenow) to SacI in the SmaI–SacI of pRS413. pREP41-gat was made a gateway destination vector by ligating the PCR-generated fragment using oligos 740 and 741 and pRS425-P_PGK as the template, cut VspI–XhoI inserted in NdeI–SalI of pREP41. The following plasmids were constructed using Gateway technology (Invitrogen): (a) pENTR-Rpa34, by a BP reaction with pDONR201 and PCR-generated fragment amplified using oligonucleotides 463 and 464 and genomic DNA from S. cerevisiae strain ODN1-1a as a template, and (b) pENTR-Sp-Rpa34 and pENTR-Sp-Rpa49, by a BP reaction with pDONR201 and PCR fragment generated using oligonucleotides 742–743 or 744–745, respectively, and genomic DNA from S. pombe strain TG11 as template. pRS425-P_PGK-CAST, pRS425-P_TEF1-CAST (Ptef1-CAST), pVV200-CAST, pEB5-CAST (YFP-CAST), pRS425-P_PGK-HA-CAST, pEB5-PAF53 (YFP-PAF53), pRS423-P_PGK-PAF53, pRS425-P_PGK-PAF53, pVV200-PAF53 (P_PGK-PAF53), pRS425-P_TEF1-PAF53 (P_TEF1-PAF53), pRS425-Sp-Rpa49, pVV200-Sp-Rpa49, pREP41-Sp-Rpa49, pEB5-Sp-Rpa34 (YFP-Sp-Rpa34), pRS425-P_PGK-Sp-Rpa34, pVV200-Sp-Rpa34, pREP41-Sp-Rpa34, pRS425-P_PGK-Rpa49 (2µ-RPA49), pGID-Rpa49, pFL36-Rpa34, pRS425-pEB5-Rpa34 (YFP-Rpa34), pGID-Rpa34, pRS425-P_PGK-Rpa34DC58 (Rpa34-ΔC), and pEB5-Rpa34DC58 (YFP-Rpa34ΔC) were obtained by LR reaction between destination vectors and entry vector and named according to vectors used. Plasmids pGEX-6P-3, pGEX-6P-3-rpa49GST, pet28a-rpa43HA, and pet28a-rpa34HA were used for recombinant expression in E. coli and are described for expression.

Complementation was tested using plasmid-shuffling assays. A null allele of haploid tester strain is complemented by the corresponding WT genes borne on URA3-containing plasmids. Fluoroorotate (FOA) is toxic for URA3+ strains (Boeke et al., 1984). FOA is used to apply a strong positive selection on cells without plasmids containing WT genes. Complementation is tested by examining whether plasmids expressing a given cDNA could bypass the lethal phenotype, monitored by FOA colonies.

GIM analysis of RPA34 deletion mutant was performed as described previously (Decourty et al., 2008). Microarray data were normalized using MATLAB (MathWorks, Inc.), and relative enrichments are described in Table S3. Genetic interactions of interest were individually confirmed by plasmid-shuffling experiments. Deletion mutants of interest from the EUROSCARF collection were mated individually with the query strain carrying RPA34 deletion tag with NAT marker under the control of the MF(ALPHA)2/YGL089C promoter (prMFα2) that allowed selectivity for MATα haploid cells (Decourty et al., 2008). The strain also contains the complementing plasmid pGID-RPA34 that allows expression of Rpa34 under the control of its own promoter terminator and carrying HPH and klURA3 markers. Diploids were selected on YPD medium complemented with hygromycin and G418. Double mutants were generated by sporulation of diploids on RSS media and selection on YPD complemented with hygromycin, G418, and ClonNat.

Immunoprecipitation of Rpa190 with coexpressed CAST and PAF53 was performed as described previously (Galy et al., 2004). In brief, yeast cells were lysed by French press in buffer A (10 mM Hepes, pH 7.5, 150 mM KCl, 1.5 mM MgCl₂, 1 mM DTT, 1 mM PMSF, 0.1% Triton X-100, 10 µg/ml leupeptin, and 10 µg/ml pepstatin). The lysates were centrifuged at 15,000 g for 30 min at 4°C. TAP-tagged Rpa190 was purified using IgG Sepharose, and bound fractions were eluted with acid elution. Isolation of ribosomes under low salt conditions by sucrose gradient centrifugation was performed as previously described (Gadal et al., 2002). In summary, cells were harvested and lysed by vortexing with glass beads in buffer containing 10 mM Tris-HCl, pH 7.5, 100 mM NaCl, 30 mM MgCl₂, and 0.35 mM cycloheximid. Nuclei and cell debris were pelleted by centrifugation at 14,000 g for 5 min at 4°C. Ribosomal subunits were separated by centrifugation for 12 h at 100,000 g in a SW40 rotor at 4°C on a 10–40% sucrose gradient prepared in the lysis buffer. Immunoprecipitation of Pol I dimers from diploid yeast strain bearing Rpa190-HA and Rpa190-Myc (BEN24) was derived from chromatin immunoprecipitation procedures. Exponentially growing cells were treated with formaldehyde 1% final concentration during 15 min. Cell were lysed using glass beads in Precellys24 (Precellys). Soluble material was discarded, and chromatin pellet was recovered after 3,500 g centrifugation for 5 min. Chromatin was resuspended, sheared using sonication, then treated with 20 U RQ1 DNase (Promega) for 100 µl of extract for 30 min at 37°C. Pol I Immunoprecipitation from such chromatin fraction was performed as described previously (Galy et al., 2004). Antibodies used were anti-HA (12CA5; Babco), anti-Myc (9E10.3; Santa Cruz Biotechnology, Inc.), anti-PAF53 (P95220; BD), and anti-Nhp2 (Henras et al., 2001). The monoclonal antibody 72B9 (provided by J. Cavaillé, University of Toulouse, Toulouse, France) was produced and characterized as an anti-fibrillarin antibody cross-reacting with the budding yeast Nop1 (Reimer et al., 1987). PFGE was performed using CHEF-DRIII (Bio-Rad Laboratories). Budding yeast chromosomes were separated in a 1% agarose gel using a 87–162 switch time ramp at an included angle of 120° during 24 h at 6 V/cm. Southern hybridization was performed using a 3.5-kb NcoI–NcoI fragment from pNOY373 (Wai et al., 2000) as an rDNA probe, labeled by random priming (Megaprime DNA labeling system; GE Healthcare).

Genes encoding Rpa43 and Rpa34 were cloned into the pET28a vector, harboring the kanamycin-resistance marker. These genes were amplified from genomic DNA with the primers 804–805 and 806–807, respectively, adding a C-terminal tag HA before inserting them into pET-28a (EMD) at site NcoI–BamHI. pGEX-6P-3 (GE Healthcare) harboring the ampicillin-resistance marker was used for GST expression. The Rpa49 gene was cloned into the pGEX-6P-3 vector, harboring the ampicillin-resistance marker and allowing N-terminal tagging with GST. For coexpression, the E. coli BL21 (Agilent Technologies) was transformed with (a) pGEX-6P-3-rpa49GST plus pet28a-rpa43HA, (b) pGEX-6P-3-rpa49GST plus pet28a-rpa34HA, (c) pGEX-6P-3 plus pet28a-rpa43HA, or (d) pGEX-6P-3 plus pet28a-rpa34HA. Transformed strains were inoculated in 0.2 liters of minimal medium plus ampicillin kanamycin with 0.01 overnight culture and grown to an OD₆₀₀ of 0.6 at 37°C, and production of recombinant proteins was induced for 3 h at 37°C with 0.1 mM isopropyl-d-thiogalactopyranoside. Purification of GST-tagged proteins was performed as described previously (Künzler and Hurt, 1998). The E. coli cell pellet was resuspended in 1% PBS buffer Triton X-100, and cells were lysed by sonication. The supernatant was incubated with 100 µl glutathione Sepharose 4B (Amersham), the beads were washed with lysis buffer, and the proteins bound to the beads were eluted with sample buffer. GST and GST-Rpa49 expression were confirmed by SDS-PAGE followed by Coomassie staining. Detection of Rpa43-HA and Rpa34-HA was achieved by Western blotting using HA peroxydase (Roche).

After centrifugation, yeast cells were resuspended in synthetic complete medium (DIFCO), mounted on a slide, and observed in the fluorescence microscope. Wide-field fluorescent images were captured with a microscope (IX-81; Olympus) equipped with a polychrome V monochromator and a camera (CoolSNAP HQ; Roper Industries) controlled with MetaMorph acquisition software (version 6; Universal Imaging Corp.). Confocal microscopy was performed using a Nipkow-disk confocal system (Revolution; Andor) installed on a microscope (IX-81), featuring a confocal spinning disk unit (CSU22; Yokogawa) and a cooled electron multiplying charge-coupled device camera (DU 888; Andor). The system was controlled using the FAST mode of Revolution IQ software (Andor). Images were acquired using a 100× Plan Apo 1.4 NA oil immersion objective and a twofold lens in the optical path. Single laser lines used for excitation were diode-pumped solid-state lasers exciting GFP fluorescence at 488 nm (50 mW; Coherent) and mCherry fluorescence at 561 nm (50 mW; Cobolt jive), and a bi-bandpass emission filter (Em01-R488/568-15; Semrock) allowed collection of the green and red fluorescence. Pixel size was 65 nm. For quantification of nucleolar volume, z stacks of 41 images with a 250-nm z step were used. Exposure time was 200 ms. Digital pictures were processed using Photoshop software (version CS3; Adobe).

For morphological analysis of nucleoli, budding yeast were cryofixed by high pressure freezing (EMPACT; Leica) and cryosubstituted with 0.02% OsO₄, 0.1% uranyl acetate, and 1% glutamate in acetone for 72 h. Cells were embedded in a Lowicryl resin (HM-20) polymerized at −50°C. Sections of 100 nm were analyzed with a 1,200× electron microscope (Jeol).

Chromatin spreading was performed as described previously with minor modifications (Osheim et al., 2009). Carbon-coated grids were made hydrophilic by glow discharge instead of ethanol treatment. Depending on the contrast of the spread chromatin, counterstaining with heavy metal can be avoided. Negatively stained chromatin was obtained by short incubation with heavy metal followed by quick drying of the sample. The positions of polymerases and rDNA fiber were determined by visual inspection of micrographs using ImageJ (National Institutes of Health). Digital pictures were processed using Photoshop.

We estimated experimental measurement errors of 2 pixels of systematic bias and 2 pixels of random error, with a pixel size of 3 nm. Random and simulated distances between adjacent polymerases were determined assuming 185 possible Pol I positions on a single rDNA gene (12 nm for each polymerase distributed along a 2.22-µm-long gene). Measurement error was added to each distance, random noise being approximated using the normal law of 5 nm of variance. Each random and simulated dataset corresponds to 100 genes with 142, 62, or 36 polymerases per gene for 25 copy, 190 copy WT, and 25 copy strains bearing rpa49Δ, respectively. Distances were generated using MATLAB. Fitting of atomic coordinates into electron microscopy density maps, modeling of two successive elongating Pol I molecules, and image creation were performed using University of California San Francisco chimera (Pettersen et al., 2004).

Table S1 lists yeast strains used in this study. Table S2 lists oligonucleotides used in this study. GIM analysis of RPA34 deletion mutant was performed as described previously (Decourty et al., 2008).

We thank Dr. Jackie Russell for comments and critical reading of the manuscript. We thank Dr. Violette Morales for providing HeLa cell extracts, A. Diot for S. pombe WT cells, Dr. Yasuhisa Nogi for pYN1003usds plasmid and rpa51Δ S. pombe strain, and E. Bertrand (pEB5) and Frederic Beckouët (pENT-Rpa34DC58) for plasmid gifts. We are grateful to Ann Beyer, Yvonne Osheim, and Sarah French for their advice on chromatin spreading. We thank M. Riva for polyclonal antibodies and Y. Henry for critical reading of the manuscript. We acknowledge Stéphanie Balor and Nacer Benmeradi for their help in electron microscopy acquisition.

This work was supported by an Action Thématique et Incitative sur Programme (ATIP) Jeunes Chercheurs grant from Centre National de la Recherche Scientifique, by Agence Nationale de la Recherche (Nucleopol and Ribeuc programme), and Jeune équipe from Fondation pour la recherche médicale (FRM). B.A. is supported by a Ph.D. fellowship from FRM. The Wellcome Trust and Cancer Research UK supported research in the laboratory of J.C.B.M. Zomerdijk. This work also benefited of the assistance of the electron microscopy facility of the Institut Fédératif de Recherche 109 and of the imaging platform of Toulouse TRI.