Transcription Termination Factor reb1p Causes Two Replication Fork Barriers at Its Cognate Sites in Fission Yeast Ribosomal DNA In Vivo

ABSTRACT Polar replication fork barriers (RFBs) near the 3′ end of the rRNA transcriptional unit are a conserved feature of ribosomal DNA (rDNA) replication in eukaryotes. In the mouse, in vivo studies indicate that the cis-acting Sal boxes required for rRNA transcription termination are also involved in replication fork blockage. On the contrary, in the budding yeast Saccharomyces cerevisiae, the rRNA transcription termination factors are not required for RFBs. Here we characterized the rDNA RFBs in the fission yeast Schizosaccharomyces pombe. S. pombe rDNA contains three closely spaced polar replication barriers named RFB1, RFB2, and RFB3 in the 3′ to 5′ order. The transcription termination protein reb1 and its two binding sites, present at the 3′ end of the coding region, were required for fork arrest at RFB2 and RFB3 in vivo. On the other hand, fork arrest at the strongest RFB1 barrier was independent of the above transcription termination factors. Therefore, RFB2 and RFB3 resemble the barriers present in the mouse rDNA, whereas RFB1 is similar to the budding yeast RFBs. These results suggest that during evolution, cis- and trans-acting factors required for rRNA transcription termination became involved in replication fork blockage also. S. pombe is suggested to be a transitional species in which both mechanisms coexist.

During eukaryotic ribosomal DNA (rDNA) replication, the fork moving opposite to transcription is arrested at replication fork barriers (RFBs) close to the 3Ј end of the coding region (3, 13, 24-27, 42, 43). RFBs must play a relevant biological role, since they are highly conserved in eukaryotes. Due to the polar nature of RFBs, rDNA is replicated mainly in a unidirectional mode cooriented with transcription. Thus, one possible role for the RFB may be to prevent the deleterious effects of head-on collisions between replication and transcription machineries (32). Since the DNA sequence at the RFB is not sufficient per se to stall replication (4,28), fork arrest must be induced by a protein factor(s) bound to the rDNA at the barrier.
In Saccharomyces cerevisiae rDNA, protein Fob1 is required for RFB activity (19), although it is still unknown whether it arrests rDNA replication by binding to the RFB sites or through a different mechanism. Functional RFBs are required for HOT1 recombination, contraction and expansion of the rDNA repeat number, and the formation of extrachromosomal ribosomal circles (17)(18)(19)(20), suggesting that RFB activity stimulates recombination occurring at the rDNA locus in this budding yeast (2). On the other hand, it has been recently shown that RFBs and HOT1 recombination are independent activities although they share cis-acting sequences (41).
In mouse rDNA, replication forks stall at the rRNA transcriptional terminator elements known as Sal boxes (27), which are the specific binding sites for transcription termination factor mTTF-1 (12,21). This protein was able to arrest replication forks in an in vitro replication assay (10,35). These in vivo and in vitro results suggest that Sal boxes and mTTF-1 block replication forks with the opposite polarity as they direct transcription termination. A protein factor that specifically binds to 27-bp repeated sequences located at the barrier has been proposed also to be involved in the RFBs of pea rDNA (28).
Contrary to what happens in the mouse, neither the rRNA transcription termination factor Reb1p nor its rDNA binding sequence seems to be involved in the RFB of S. cerevisiae (4,41). These observations suggest that the molecular mechanism that regulates rDNA replication arrest diverged through evolution. In the present work, we found three independent closely spaced RFBs in the fission yeast rDNA. Two of these RFBs required both the transcription termination factor reb1p and its two binding sites near the 3Ј end of the 25S gene, whereas the other RFB functioned in the absence of these cisand trans-acting factors. Therefore, Schizosaccharomyces pombe could be a transitional species in which the mechanisms operating in budding yeast and mammals coexist.
Construction of plasmids containing rDNA sequences. Plasmid pBL1263, containing a complete repeat of S. pombe rDNA (23), was the source of the sequences analyzed for RFB activity. Autonomously replicating plasmid pIRT2 (14) was used as a vector. For convenience, the HindIII insert in pBL1263 was first cloned into the HindIII site of pUC18 generating plasmid pHH10.4. Plasmids pIRT1.6(ϩ) and pIRT1.6(Ϫ) were obtained by inserting the blunt-ended 1.6-kb XhoI-HindIII rDNA fragment from pHH10.4 into the SmaI site of pIRT2. The (ϩ) plasmids contained the insert in the orientation expected to block fork progression. Construction of plasmid pRebs was carried out as follows. A PCR product was obtained using primer 1 (5Ј-CCCCTGCAGTTTTGAAGAGATA AAAGG-3Ј) and primer 2 (5Ј-CCCGGATCCTTTTACTAGGATTTGTGC-3Ј) on pIRT1.6(ϩ). Each primer contained 18 nucleotides (underlined) that annealed a few nucleotides upstream from the 5Ј reb1p binding sequence (primer 1) or downstream from the 3Ј reb1p binding sequence (primer 2), a PstI (primer 1) or BamHI (primer 2) restriction site and a CCC tail. This fragment was PstIϩBamHI digested and inserted into the polycloning site of pIRT2 near ars1. Thus, the 229-bp fragment cloned in pRebs contained, in addition to both reb1p binding sites and the sequence between them, 15 bp upstream from the 5Ј binding sequence and 14 bp downstream from the 3Ј binding sequence. For construction of plasmids pRebBS(ϩ) and pRebBS(Ϫ), a double-stranded oligonucleotide containing the 17-bp reb1p binding site (5Ј-AGGTAAGGGTAATG CAC-3Ј) (44) was inserted in both orientations into the PstI and BamHI sites of the pIRT2 polycloning site. To direct the ligation to the desired orientation, the oligonucleotide contained PstI and BamHI sticky ends. Appropriate insertion was checked by sequencing. For the construction of plasmid pRebs⌬BS, a PCR product was obtained by using oligonucleotides 5Ј-AAGGCCTAAATCCTAGT AAAAGGATC-3Ј and 5Ј-AAGGCCTTTTCCCTTCAAAAAG-3Ј. These primers annealed divergently next to each side of the reb1p binding sequence closer to ars1 in pRebs (underlined nucleotides) and contained a StuI site at their 5Ј ends (boldface nucleotides). After digestion with StuI, the PCR product was ligase-mediated circularized. For the construction of pRebsSEP, first a StuI site was created by PCR between both reb1p binding sequences of plasmid pRebs. This StuI site was located 60 bp from the binding site closer to ars1 and at 106 bp from the other binding site. Then, the 787-bp EcoRV-NruI fragment from the pBR322 tetracycline resistance gene was cloned into this StuI site.
Deletion of reb1 ؉ gene. Deletion of the reb1 ϩ gene was achieved by PCRmediated replacement of the complete open reading frame by the kanMX6 ϩ gene (1). PCRs were performed with two 84-nucleotide-long primers (5Ј-GAT ATTAGCG ATTGATAAGT TGAAGTGATT ACTCAATTAT AGTACT TCAA AAATATAATC CGCCAGGGTT TTCCCAGTCA CGAC-3Ј and 5Ј-A TTGTAAGGA CGTCAATTGG AGAATCCAGA AAGTACCACT TTAAA GTCAT CAATGGCTGA AGCGGATAAC AATTTCACAC AGGA-3Ј), where the first 60 nucleotides (underlined) corresponded to sequences flanking the reb1 ϩ open reading frame. The remaining 24 nucleotides of each primer corresponded to sequences located at either side of the pBluescript SK(ϩ) polycloning site, where kanMX6 ϩ was cloned (a gift from S. Moreno). Transformation of S. pombe 117ϫ118 diploid strain with the PCR fragment and selection of G418-resistant diploids were performed as described previously (1). Genomic DNA from selected transformants was digested with EcoRV, electrophoresed, and hybridized with a probe specific for kanMX6 ϩ gene to confirm its integration. reb1 ϩ /reb1⌬ diploids were induced to form spores to allow further analysis by tetrad separation. The four spores from selected asci were again checked for integration of the kanMX6 ϩ gene as described above (see Fig. 5C).
Two-dimensional gel electrophoresis. Genomic and plasmid DNAs for twodimensional (2D) gel analysis were isolated from asynchronous log phase cultures by using the procedures described by Caddle and Calos (5) (for plasmid analysis) or by Huberman and coworkers (15) (for genomic rDNA analysis). Electrophoresis conditions were as described in reference 9.

RESULTS
S. pombe rDNA contains three closely spaced RFBs. The S. pombe genome contains ϳ100 copies of rRNA genes organized in two arrays near both ends of chromosome III (29,33). Two 17-bp binding sequences for the transcription termination factor reb1p, separated by 166 bp, are present in the nontranscribed spacer (NTS) close to the 3Ј end of the 25S gene ( Fig.  1) (44). Replication of S. pombe rDNA was analyzed by 2D agarose gel electrophoresis. DNA from exponentially growing strain 972 h Ϫ cells was digested with the restriction enzymes indicated, separated in 2D gels, transferred, and hybridized with probes specific for a series of overlapping fragments covering the rDNA repeat ( Fig. 1, fragments A through E). Analysis of fragment A showed a spot corresponding to an accumulated Y-shaped replication intermediate (Fig. 1A, arrow), confirming previous observations of a replication barrier in S. pombe rDNA (26,37). However, the elongated appearance of this signal suggested that forks stalled at several sites rather than at a single site. 2D gel analysis of the overlapping fragment B supported this possibility. Three independent spots were identified on the descending portion of the simple-Y arc (Fig. 1B), indicating that replication stalled at three alternative sites, herein called RFB1, RFB2, and RFB3. Equivalent spots on the ascending portion of the arc were absent, indicating that, as in other species, these three RFBs were polar, arresting only forks moving against the direction of transcription (leftwards in the map of Fig. 1). The intensities of the spots of accumulated replication intermediates were clearly different. The strongest signal corresponded to the first pausing site that leftward moving forks encounter (RFB1). The middle site, RFB2, gave the weakest signal, and RFB3 produced a signal of intermediate intensity. Replication analysis of fragment C (HindIII-SalI) showed no spots of accumulated intermediates (Fig. 1C), indicating that DNA sequences involved in fork stalling were located to the left of the HindIII site. Two additional fragments, corresponding to the coding region, revealed no replication impediment, as uniform simple-Y arcs were observed ( Fig. 1D and E).
A bubble arc was also visible in the 2D gel corresponding to fragment C (Fig. 1C), indicating that a replication origin located within this fragment fires in a fraction of rDNA repeats. This observation is in agreement with the previous identification of an autonomously replicating sequence (ARS) (ars3001) in the NTS of S. pombe rDNA (37).
All three S. pombe RFBs are active in autonomously replicating plasmids with the same polarity and relative efficiency as in the chromosome. As described above, S. pombe RFBs are located in the NTS, 5Ј to the HindIII site. Therefore, the 1.6-kb XhoI-HindIII restriction fragment of S. pombe rDNA ( Fig. 2A) was cloned in both orientations close to the ars1 replication origin of vector pIRT2 (Fig. 2B). This fragment contained the NTS portion lying to the left of the HindIII site and the 3Ј end of the 25S gene. In pIRT1.6(ϩ) the inserted sequence is replicated in the direction in which the barriers are active in the chromosome, whereas in pIRT1.6(Ϫ) the insert is replicated in the opposite direction. Replication of these plasmids in strain 35 was analyzed by 2D gel electrophoresis after double digestion with PvuII and EcoRV, using a specific probe to detect the fragment containing the insert. Locating the insert close to ars1 ensured that the clockwise-moving fork would reach the RFBs before the fork moving counterclockwise entered the fragment analyzed. If the RFBs were active, simple-Y-shaped intermediates of this fragment would accumulate while the counterclockwise fork replicated the other fragment. The results obtained are shown in Fig. 2C and D. All three RFBs observed in the chromosomal context were also detected in pIRT1.6(ϩ), visualized as three spots of accumulated intermediates on the Y arc (Fig. 2C). The relative intensities of these spots fitted well with those observed in the chromosome. No arrest sites were detected in pIRT1.6(Ϫ) (Fig. 2D), indicating that all VOL. 24, 2004 rDNA REPLICATION FORK BARRIERS IN S. POMBE three RFBs retained the same polarity as in the chromosomal context. No additional barriers were detected within the rightward HindIII-BamHI fragment of the NTS (not shown). A longer exposure of the same gel as that shown in Fig. 2C allowed detection of a partial bubble arc (Fig. 2C'), generated upon bidirectional replication from ars1. In addition, a straight line emerged from the spots of accumulated intermediates extending upward and to the left in a diagonal fashion (Fig. 2C'). This signal corresponded to double-Y intermediates generated when the counterclockwise-advancing fork entered the fragment until it encountered the clockwise fork arrested at the barriers. These observations showed that in a significant fraction of the plasmid molecules replication termination occurred at the barriers.
Location of RFB1. The spots of accumulated intermediates showed in Fig. 2 appeared on the simple-Y arc to the right of the inflexion point. This indicated that RFBs were located closer to the HindIII site. Therefore, to locate the DNA sequence required for RFB1, we analyzed a fragment spanning 383 bp next to the HindIII site (Fig. 3A). This fragment was cloned into pIRT2 and two restriction fragments of the resulting plasmid (p3ЈRebs) (Fig. 3B) were analyzed. 2D gels showed a single strong spot of a specific Y-shaped intermediate corresponding to RFB1 (Fig. 3C and D). Replication termina-tion structures were also visible ( Fig. 3C and D), indicating termination at the barrier.
The 17-bp binding sequence for the rRNA transcription terminator factor reb1p is required and sufficient to induce replication fork arrest at RFB2 and RFB3. As mentioned before, S. pombe rDNA contains two identical 17-bp binding sequences for the transcription termination protein reb1p (30,44). Both termination signals are included in the fragment where RFBs were mapped (Fig. 2). To address if these binding sequences are also DNA cis-acting elements for the remaining barriers RFB2 and RFB3, we tested the capacity of a 229-bp fragment, containing both reb1p binding sites (Fig. 4A), to arrest replication forks. This fragment was cloned in the proper orientation into pIRT2 (pRebs) (Fig. 4B). The 2D gel of the PvuII-BstEII fragment containing the insert showed two signals of accumulated intermediates at the expected positions on the simple-Y arc (Fig. 4D), indicating that this fragment contains the cis-acting signals required for RFB2 and RFB3. We removed one of the 17-bp binding sequences (the one closer to ars1) from this plasmid (the modified insert is shown in Fig. 4B). As a consequence of this deletion, one of the spots was missing in the 2D gel of the resulting plasmid pRebs⌬BS and only the spot closer to the inflexion of the Y arc remained (Fig. 4E). These results indicated that reb1p binding sites are essential cis-acting signals for RFB2 and RFB3 and that they function independently. When both binding sites were placed further apart by inserting a 787-bp sequence between them (Fig. 4C), the two fork arrest positions also appeared to be separated from each other as indicated by the new relative locations of the spots in the corresponding 2D gel (compare panels F and D of Fig. 4). In addition, we found that the 17-bp sequence was not only necessary but also sufficient for fork arrest. A synthetic sequence identical to the reb1p binding site inserted into pIRT2 (Fig. 4G) induced fork arrest in the (ϩ) orientation (Fig. 4H) and had no effect in the opposite (Ϫ) orientation (Fig. 4I). Deletion of reb1 ؉ abolishes fork arrest at RFB2 and RFB3. Since the reb1p binding sequence was necessary and sufficient to induce polar replication fork arrest, we regarded this transcription termination protein as a candidate to be involved in RFB2 and RFB3, even though the S. cerevisiae homologue Reb1p is not involved in rDNA RFBs (41).
To analyze the function of S. pombe reb1p in RFB activity, we constructed a heterozygous reb1 ϩ /reb1⌬ diploid strain. The null allele was obtained by PCR-mediated gene replacement of reb1 ϩ by the selectable gene marker kanMX6 ϩ , which confers resistance to the antibiotic G418 (Fig. 5A). Dissections of three representative tetrads of the G418-resistant reb1 ϩ /reb1⌬ diploids are shown in Fig. 5B. In the absence of G418 all four spores were viable, indicating that reb1 ϩ is not an essential gene in S. pombe (Fig. 5B, left panel). However, two spores of each tetrad gave colonies slightly smaller than the other two. In all cases the smaller colonies were those generated by reb1⌬::kanMX6 ϩ spores, as they grew in the presence of G418 (Fig. 5B, right panel). Replacement of reb1 ϩ by kanMX6 ϩ was confirmed by Southern blotting (Fig. 5C) and PCR analysis (not shown). Therefore, deletion of reb1 ϩ gene was not lethal, but mutant cells grew somewhat more slowly than wild-type cells. This small effect of reb1 ϩ deletion on cell growth was confirmed by dilution assays (data not shown).
To address the involvement of reb1p in the barriers, replication of pIRT1.6(ϩ) containing all three RFBs was analyzed in isogenic reb1 ϩ and reb1⌬ haploid strains. As shown in Fig.  5D and E, in the reb1⌬ strain the spots corresponding to RFB2 and RFB3 disappeared and only the one generated by RFB1 remained. Similar results were obtained upon analysis of plasmids pRebs and pRebBS(ϩ) (data not shown). This observation demonstrates that the transcription termination factor reb1p is required for RFB2 and RFB3. Colonies were then replicated onto a new plate containing 100 g of G418/ml (right panel) to determine the segregation of the alleles. (C) Southern blot verifying the deletion of reb1 ϩ gene by replacement with kanMX6 ϩ . EcoRV-digested DNA from reb1⌬ and wild-type haploid cells was hybridized with a probe specific for kanMX6 ϩ (left panel) or reb1 ϩ (right panel). The kanMX6 ϩ probe hybridized to the expected 2.1-kb restriction fragment from reb1⌬ DNA, containing most of the kanMX6 ϩ gene (see bottom map in panel A) and did not hybridize to the DNA from wild-type cells. As expected, the reb1 ϩ probe did not hybridize to the DNA from reb1⌬ cells but detected the 3.9-kb fragment containing reb1 ϩ in the DNA from wild-type cells. (D and E) 2D gels of plasmid pIRT1.6(ϩ) replicating in wild-type (wt) and reb1⌬ cells, respectively. The restriction fragment analyzed was the same as in Fig. 2C. Arrowheads labeled 1 to 3 point to the spots of accumulated replication intermediates induced by the three barriers in wt cells. Note that in reb1⌬ cells only the spot generated by RFB1 remained. The probe used in these autoradiograms was the same as for Fig. 2.

DISCUSSION
In S. pombe, the rRNA transcription termination factor reb1p has been recently identified, showing sequence similarity with S. cerevisiae Reb1p and mouse m-TTF1 (44). These three termination factors share myb-like DNA binding domains. S. pombe reb1p has two identical 17-bp binding sites that block read-through transcription in vitro (44). In addition, reb1p also causes in vitro 3Ј-end RNA formation at two sites of S. pombe rDNA that correspond to the transcription termination sites determined in vivo (38,44).
In this work we have identified three RFBs in the S. pombe rDNA. Fork arrest at two of these RFBs (RFB2 and RFB3) is produced upon binding of the transcription termination protein reb1, in a fashion similar to what has been proposed to occur in mouse (27,35). Removal of the reb1p binding sequence or deletion of reb1 ϩ completely abolished replication blockage at RFB2 and RFB3 (Fig. 4E and 5E). Therefore, upon binding to its cognate sequence, reb1p inhibits both rRNA transcription and rDNA replication occurring with opposite directions, thus preventing head-on collision of both machineries. It cannot be ruled out that other factor(s) besides reb1p are also required for fork arrest at RFB2 and RFB3. On the other hand, fork arrest at RFB1 is independent of this transcription termination factor, since it remains active in a reb1⌬ mutant strain (Fig. 5E). Moreover, we have found a reb1p-unrelated protein that specifically binds to a short sequence within the RFB1-containing fragment analyzed in Fig.  3 (unpublished data). Thus, S. pombe RFB1 seems to be similar to the two barriers found in budding yeast, as both of them are normal in strains with a temperature-sensitive allele of REB1 growing at restrictive temperature (41). Therefore, two different and independent mechanisms operate in S. pombe rDNA to arrest replication forks. Remarkably, although different trans-and cis-acting factors are involved in these two mechanisms, both of them block replication progression in a polar fashion. This indicates that polarity of rDNA barriers is an essential feature in accomplishing their biological role, which may be to avoid or regulate head-on collision between transcription and replication.
Besides the RFBs present in the rDNA, another barrier, named RTS1, has been described for S. pombe. RTS1 is involved in the mating-type switching by determining the direction of replication at the mat1 locus (6, 7). As in the case of rDNA RFBs, RTS1 is a polar barrier that contains a cluster of three full-length and one truncated ϳ60-bp imperfect direct repeats (7). Interestingly, each of these repeats includes a sequence that shows homology to the 17-bp reb1p binding sequence required for RFB2 and RFB3. Moreover, according to our findings, the orientation of these homologous sequences is in agreement with the reported polarity of RTS1 (7). This observation raises the possibility that reb1p also plays a role in fork arrest at the mat1 locus.
It is interesting that whereas REB1 is an essential gene in S. cerevisiae (16), deletion of reb1 ϩ in S. pombe has only a weak effect on cell growth ( Fig. 5B and D). Besides its function in RNA polymerase I transcription termination (22,36), S. cerevisiae Reb1p regulates the expression of several unrelated RNA polymerase II transcribed genes (see reference 40 and references therein). In addition, Reb1p binding sites are also present in the subtelomeric X and YЈ regions in budding yeast (8). Thus, the essential nature of Reb1p in S. cerevisiae could be due to the function(s) that this protein performs at these additional sites, but not at the rDNA locus. In agreement with this hypothesis, it was recently shown that deletion of the Reb1p binding site in all rDNA chromosomal repeats of S. cerevisiae has no effect on cell growth or rRNA synthesis (39). Altogether, these results suggest that, at least in these two species, in the absence of the cis-or trans-acting factors currently known to be involved in rRNA transcription termination, rRNA transcripts are terminated by an alternative unknown pathway and processed properly to form functional ribosomes. It would be worthy to investigate if factors involved in S. pombe RFB1 play a role in this alternative pathway.