Binding of the Replication Terminator Protein Fob1p to the Ter Sites of Yeast Causes Polar Fork Arrest*

Fob1p protein has been implicated in the termination of replication forks at the two tandem termini present in the non-transcribed spacer region located between the sequences encoding the 35 S and the 5 S RNAs of Saccharomyces cerevisiae. However, the biochemistry and mode of action of this protein were previously unknown. We have purified the Fob1p protein to near-homogeneity, and we developed a novel technique to show that it binds specifically to the Ter1 and Ter2 sequences. Interestingly, the two sequences share no detectable homology. We present two lines of evidence showing that the interaction of the Fob1p with the Ter sites causes replication termination. First, a mutant of FOB1, L104S that significantly reduced the binding of the mutant form of the protein to the tandem Ter sites, also failed to promote replication termination in vivo. The mutant did not diminish nucleolar transport, and interaction of the mutant form of Fob1p with itself and with another protein encoded in the locus YDR026C suggested that the mutation did not cause global misfolding of the protein. Second, DNA site mutations in the Ter sequences that separately and specifically abolished replication fork arrest at Ter1 or Ter2 also eliminated sequence-specific binding of the Fob1p to the two sites. The work presented here definitively established Ter DNA-Fob1p interaction as an important step in fork arrest.

Fob1p protein has been implicated in the termination of replication forks at the two tandem termini present in the non-transcribed spacer region located between the sequences encoding the 35 S and the 5 S RNAs of Saccharomyces cerevisiae. However, the biochemistry and mode of action of this protein were previously unknown. We have purified the Fob1p protein to near-homogeneity, and we developed a novel technique to show that it binds specifically to the Ter1 and Ter2 sequences. Interestingly, the two sequences share no detectable homology. We present two lines of evidence showing that the interaction of the Fob1p with the Ter sites causes replication termination. First, a mutant of FOB1, L104S that significantly reduced the binding of the mutant form of the protein to the tandem Ter sites, also failed to promote replication termination in vivo. The mutant did not diminish nucleolar transport, and interaction of the mutant form of Fob1p with itself and with another protein encoded in the locus YDR026C suggested that the mutation did not cause global misfolding of the protein. Second, DNA site mutations in the Ter sequences that separately and specifically abolished replication fork arrest at Ter1 or Ter2 also eliminated sequence-specific binding of the Fob1p to the two sites. The work presented here definitively established Ter DNA-Fob1p interaction as an important step in fork arrest.
DNA replication in many prokaryotes is often terminated at sequence-specific replication termini (Ter sites). Replication terminator proteins specifically bind to the Ter sites and arrest forks in an orientation-specific mode with respect to the replication origins (1). In many eukaryotes, although every replicon in the multiorigin chromosomes does not have specific Ter sites, such sites are present in the nontranscribed spacer sequences in the rDNA from yeast to man (1,2).
In Escherichia coli and Bacillus subtilis, the Ter sites bind to replication termination proteins, called Tus and RTP, respectively, that are contrahelicases (3,4) and impede replication fork movement not only by binding to the Ter sites but also by protein-protein interaction between the helicase and the terminator protein (5). The rDNAs of Saccharomyces cerevisiae (6,7), Schizosaccharomyces pombe (8), Xenopus (9), mouse (10), pea (11), and humans (12) have similar replication fork arrest systems located in their nontranscribed spacer elements. In S. pombe, replication fork arrest has been shown to occur at and near the mating type switch locus (13)(14)(15).
The rDNA locus of S. cerevisiae ( Fig. 1) consists of 100 -200 tandem copies of 9.1-kb rDNA units present in chromosome number XII. Each unit of the rDNA consists of a 35 S rRNA and a 5 S rRNA gene that transcribe in opposite directions. A nontranscribed spacer called NTS2 containing an origin of replication called ARS 1 (autonomously replicating sequence) is located between the genes encoding the 35 S and the 5 S RNAs (6,16).
The two replication forks initiated at the ARS face unequal fate. The rightward moving fork moves through the 35 S rDNA in the same direction as that of transcription until it meets the fork coming from the opposite direction. But the leftward moving fork, after passing through the 5 S rRNA gene, is arrested at two Ter sites located in the nontranscribed spacer called NTS1 and is thus prevented from entering the 35 S rRNA gene. The two Ter sites (known as replication fork barrier or RFB sites) located in NTS1 arrest replication forks in an orientationdependent manner (6,7). The Ter1 and Ter2 sites are separated from each other by a few nucleotides and can be further separated by inserting foreign DNA between them without affecting their activities (7).
The Ter sites in prokaryotes and eukaryotes are recombinogenic (2). The yeast rDNA also contains a recombination hot spot HOT1 that consists of two elements called enhancer and initiator that include the RNA polymerase I enhancer and promoters, respectively (see Fig. 1). The enhancer region includes the tandem Ter sites. The initiator and enhancer elements together stimulate high levels of recombination between tandemly repeated sequences even when placed at ectopic sites in a chromosome (17)(18)(19)(20).
Deletions of Ter sites (present in the enhancer element) and/or that of the promoter (present in the initiator element), as expected, reduce not only the HOT1 activity (18,21) but also fork arrest (22). A colony sectoring assay has been developed to identify mutations in genes that reduce HOT1 activity, and this technique has resulted in the identification of several genes that are involved in the recombination process (19,21). The FOB1 gene was discovered using such a colony sectoring assay and was shown by two-dimensional gel analysis to be involved in fork arrest at the Ter sites of rDNA (22). FOB1 also plays a role in rDNA circle formation and aging (23) and in expansion and contraction of rDNA repeats (24 -26).
Although several functions of FOB1 have been described by in vivo experiments, very little is known about its biochemical properties and its precise mechanistic role in promoting repli-cation termination. In the present work, we describe the isolation of several interesting mutants of FOB1 and purification of the wt and the mutant forms of the protein expressed in yeast to near-homogeneity. We show further that Fob1p protein specifically binds to the Ter1 and Ter2 sites. We present two lines of evidence in support of the proposition that interaction of Fob1p with the Ter1 and Ter2 sites causes replication termination. First, we show that mutation in the FOB1 locus that significantly reduces the DNA binding activity of the encoded protein also abolishes the ability of the protein to arrest replication forks in vivo. Second, mutations at Ter1 and Ter2 sites that separately inactivate replication termination at these sites also abolish binding of Fob1p to the two sequences. From the results, we conclude that interaction of Fob1p with Ter sites creates a protein-DNA complex that arrests replication forks in an orientation-specific fashion with respect to the nearest ARS site.
Plasmids-For two-hybrid analysis, FOB1 and its mutants were cloned as BamHI-SalI fragments into pGAD424 and pGBT9 (27). For sectoring/ papillae formation assay, the LEU2 marker in pGAD424 and pGAD424 containing the FOB1 gene (GAD-FOB) or mutants was replaced by a hygromycin marker by gap repair cloning generating pGADh and its derivatives. For tandem affinity purification (TAP) tagging, the fragment containing ProtA-TEV-CBP site was amplified by PCR from the plasmid pBS1761 (28) and cloned as a BamHI fragment at the 5Ј end of the FOB1 gene in GADhFOB plasmid (TAP-FOB). Fragment containing the glutathione S-transferase (GST) gene under the Gal1 promoter and the HIS3 marker was amplified by PCR (29) and was used to construct the tripletagged FOB1 expression plasmid GST-TAP-FOB.
Sectoring Assay-Colony color sectoring assay was performed as described (21). The vector pGAD424 or pGADh plasmid and their derivatives containing the wt FOB1 gene or its mutants were transformed into ⌬fob1H strain and plated either on SD/Ura Ϫ Leu Ϫ low Ade containing plates or on YPD ϩ hygromycin plates. On YPD ϩ hygromycin plates, growth was much faster, and sectoring/papillae formation was distinct and more frequent than on the plates containing minimal medium.
Mutagenesis-Mutagenesis of FOB1 gene was carried out by PCR with Taq polymerase with GAD forward and reverse primers in the presence of MnCl 2 and low dNTP (one dNTP) concentrations. The mutants were co-transformed with a SmaI-or BamHI-digested pGAD424 or pGADh plasmid into ⌬fob1H strain and plated either on SD/Leu Ϫ Ura Ϫ low Ade or on YPD ϩ hygromycin plates, respectively. The colonies that failed to sector were further streaked on appropriate plates to confirm the non-sectoring phenotype. Plasmids were isolated from the nonsectoring clones and transformed into E. coli to amplify and recover the plasmid DNA. The plasmids isolated from E. coli were again transformed into ⌬fob1H to confirm the loss of sectoring and then sequenced to identify the mutations.
FOB1 Deletions-FOB1 deletion was carried out as indicated below. FOB1 gene was obtained from genomic DNA by PCR using Vent polymerase (New England Biolabs) and was cloned into pUC18 at the SmaI site. pUC18-FOB1 plasmid was digested with BstZ17I or SnaBI and BstZ17I and ligated to the G418 cassette amplified by PCR from the plasmid pFA6KnMX4 (30). The fragment containing the G418 cassette flanked by the ends of FOB1 was amplified by PCR with pUC forward and reverse primers and transformed into the strain HRM1-1 or SUB62 to delete the FOB1 gene. FOB1 deletion was confirmed by PCR and by two-dimensional gel analysis to confirm loss of replication fork blockage activity.
Two-hybrid Assay-For two-hybrid assay, the pGBT9 and pGAD424 plasmids and their derivatives containing FOB1 gene, its mutants, or deletions were transformed sequentially into the yeast strain PJ69-4A (31) and patched on SD/Leu Ϫ Trp Ϫ , SD/Leu Ϫ Trp Ϫ Ade Ϫ , or SD/ Leu Ϫ Trp Ϫ His Ϫ (with 2 mM aminotriazole) plates, and the interactions were subsequently confirmed by liquid ␤-galactosidase assay (described in Clontech manual).
DNA Preparation and Two-dimensional Gel Analysis-Replication intermediate DNA was prepared from yeast cells growing in appropriate medium to A 600 0.8 -0.9. Cell lysis, DNA preparation, and twodimensional gel analysis were carried out as described previously (32). The DNA was transferred to Nytran membrane (Schleicher & Schü ll), and hybridization was carried out as described (33).
Purification of GST-TAP-Fob1p Protein-An overnight culture of ⌬fob1SUB62 cells containing the GST-TAP-FOB1 plasmid was inoculated into 10 liters of YPD with 2% raffinose and 200 g/ml hygromycin. Galactose was added to 2% final concentration at midlog phase, and the cells were harvested after 5 h and stored at Ϫ70°C. Cells were quickly thawed and suspended in 100 ml of buffer A (10 mM HEPES, pH 7.9, 10 mM KCl, 0.5 mM MgCl 2 , 0.5 mM dithiothreitol, 2 mM benzamidine, 0.5 mM phenylmethylsulfonyl fluoride, a mixture of leupeptin, pepstatin, and aprotinin). Acid-washed glass beads (half the total volume of buffer A) were added to the cells, and the cells were lysed in a Waring Blendor by setting the blender speed to 7 and running the blender 6 times for 30 s each with 4-min intervals between each run. KCl was added to 150 mM final concentration. The lysate was first centrifuged in a Sorvall RC5B HS4 rotor at 5000 rpm for 10 min and then centrifuged in a Beckman ultracentrifuge at 30,000 rpm for 20 min in a Ti70 rotor, and the supernatant (fraction I) was collected. Ammonium sulfate was added to the fraction I to 70% final concentration, and it was centrifuged in the ultracentrifuge Ti70 rotor at 20,000 rpm for 20 min. The ammonium sulfate pellet was suspended in (150 ml) buffer B (10 mM Tris⅐HCl, pH 8.0, 5% glycerol, 10 mM ␤-mercaptoethanol, 1 mM MgCl 2 , 1 mM imidazole, 2 mM CaCl 2 , 2 mM benzamidine, 0.5 mM phenylmethylsulfonyl fluoride, a mixture of leupeptin, pepstatin, and aprotinin), and conductivity was adjusted to that of buffer C (buffer B ϩ 150 mM NaCl). 5 ml of calmodulin beads (Stratagene) were washed with buffer C, and the lysate was bound to the calmodulin beads at 4°C for 2 h. The slurry was poured into a column, and the beads were washed with 100 ml of buffer C followed by 50 ml of buffer D (buffer B ϩ 0.65 M NaCl). The protein was eluted with 5 fractions of 5 ml each of buffer E (buffer C with 2 mM EGTA instead of 2 mM CaCl 2 ). The last four fractions containing the Fob1p fusion protein were pooled and loaded on a 1-ml Mono Q column in a fast pressure liquid chromatograph. The column was washed with 3 ml of buffer C and eluted with buffer C containing a gradient of 150 mM to 1.5 M NaCl. The fractions containing Fob1p protein were pooled and bound to 2.5 ml of glutathione-Sepharose beads (Amersham Biosciences). The beads were washed with 50 ml of buffer C and 50 ml of buffer F (buffer B ϩ 0.8 M NaCl). Fob1p protein was eluted with buffer C containing 50 mM Tris⅐HCl, pH 8.0, and 10 mM reduced glutathione. The protein, which was in a volume of 12.5 ml, was further concentrated by binding to 1 ml of calmodulin beads, washing with 10 ml of buffer B, followed by washing with 10 ml of buffer F and eluting with 10 0.5-ml fractions of buffer E. Fractions 2 and 3 were used for all experiments.
Purification of TAP-Fob1p Protein for DNA Binding Assay-The SUB62 cells expressing TAP-Fob1p fusion protein were grown in 500 ml of medium and harvested. The cells were lysed in a bead beater and centrifuged as above. The supernatant was bound to 200 l (packed volume) of calmodulin beads as above in buffer C and washed with buffer C. After washing, 20 -50 l of beads were used for DNA binding as described below. For some experiments we have made large scale preparations of this TAP-Fob1p protein with some modifications that will be published elsewhere.
DNA for Fob1p Binding Assay-pUC18 plasmid DNA was digested with HaeIII or DdeI, and the digestion products were dephosphorylated by shrimp alkaline phosphate enzyme (U. S. Biochemical Corp.). An ϳ370-bp-long Ter fragment was generated by PCR with RFBMAP1 and RFBMAP2 primers, the sequences of which are available upon request. The pUC18 fragments and the Ter (RFB) fragment were end-labeled using [␥-32 P]ATP and T4 polynucleotide kinase. The fragments were used separately and mixed together also (see Fig. 8). Custom-made oligonucleotides of different lengths representing TerI, TerII, and IR (inverted repeat) and their mutant forms were labeled as above. Briefly, 2 pmol of each oligonucleotide was labeled and mixed with 4 pmol of the complementary oligonucleotide in the presence of 10 mM Tris⅐HCl, pH 8.0, 1 mM EDTA, pH 8.0, and 600 mM NaCl. The mixture was heated in boiling water for 3 min and cooled for several hours (typically overnight). All the labeled DNA fragments were passed through a Sephadex G-25 spin column before use. Oligonucleotides of various lengths have been used in these experiments. Although most of the sequences are described in the figures and text, detailed sequences are available upon request.
DNA Binding Assay-To the TAP-Fob1p protein bound to calmodulin beads, 20 -50 fmol of labeled DNA was added in the presence of 5 g of poly(dI-dC) in a volume of 200 l of buffer C (calmodulin binding buffer), and the binding was carried out at room temperature for 15 min. The reaction mixture was loaded on a Bio-Rad polyprep column (catalogue 731-1550), and the beads were washed with 4ϫ 1 ml of buffer C after which the protein-DNA complex was eluted with 4ϫ 100 ml of buffer E (EGTA buffer). 200 l of the eluate was mixed with DNA loading dye without or with 0.25% SDS and fractionated on a polyacrylamide gel and autoradiographed.
Fluorescence Microscopy-Cells grown in raffinose and galactose were spotted on slides, treated with DAPI for several minutes, and directly visualized by a Nikon microscope with visible, UV, and fluorescence filters. A Spot RT color camera (Diagnostic Instruments, Inc.) was used to take the photomicrographs. Fig. 1. The EcoRI-HpaI fragment should contain both of the Ter sites, and we confirmed this and calibrated our two-dimensional gel technique by cloning the hygromycin resistance marker into the pBB3NTS plasmid (7) (resulting in the plasmid pBB3-H; Fig. 2A). Both the fob1 and wild type yeast cells were transformed with the plasmid, and cultures were grown in rich medium containing hygromycin, and the plasmid DNA was extracted as described above. A typical two-dimensional gel of the SspI cut plasmid that was probed with 32 P-labeled NTS1 DNA is shown in Fig. 2C. The interpretation of the pattern is shown in diagrammatic form in Fig. 2B. Consistent with published observations, two tandem termination spots, Ter1 and Ter2, were observed on the Y arc. Most of the forks seemed to be getting arrested at Ter1 and any that leaked through were apparently arrested at Ter2. In addition to the twin Ter spots on the Y arc, two corresponding spots were observed in the X arc. The latter spots are generated by progression of forks initiated from the ARS, moving coun-terclockwise which merged with the forks that were initiated from the ARS and moving clockwise and arrested at Ter1 and Ter2, thus generating X-shaped intermediates. Previous work had shown that although all of Ter1 was located entirely within the HindIII and HpaI fragment ( Fig. 1 center, H to Hp), Ter2 consisted of sequences located not only in this fragment but also of an inverted repeat sequence located just upstream of the HindIII site (see Fig. 1, lower, marked IR) (7). The poly(dT) region located immediately upstream of the IR sequence is needed for enhanced activity of Ter2 (7). The mutations C20 and C26 specifically knock out Ter1, and the N35 mutation located in the inverted repeat knocks out Ter2 but not Ter1. Experiments utilizing some of these mutations are described below.

Replication Fork Arrest at Ter Sites-The locations and the landmarks around the replication termini derived from published works is shown in
HOT1 Induced Recombination and Isolation of fob1 Mutants-We wished to investigate the possible role of Fob1p in replication termination by endeavoring to isolate a series of mutants using the HOT1 recombination assay described (18,21). The principle of the HOT1 assay is diagrammatically shown in Fig.  3A. A yeast strain constructed by Dr. R. Keil (20,21) contained chromosomally integrated tandem alleles of leu2 and a HOT1 element inserted into the left leu2 sequence. In between the tandem leu2 sequences is a reporter sequence ADE5 whose presence in the ade2, ade5 background of the host generated a red pigment. Loss of the ADE5 marker generated white sectors. We deleted the FOB1 gene from this host by one-step gene disruption as described above and introduced a plasmid expressing FOB1 or a blank plasmid into the host. In the presence of functional FOB1, the HOT1 sequence promoted recombination between the tandem leu2 sequences resulting in a recombination product excised as a circle and containing the ADE5 marker but without a replication origin. The circle was therefore eliminated from the cell generating white sectors (in red colonies) due to the loss of the ADE5 marker (Fig. 3B, panel The NTS2 has an origin of replication called ARS. The NTS1 has two Ter sites that arrest replication fork coming from the NTS2 region in an orientation-dependent manner. The middle panel shows an expanded view of the NTS1 area containing Ter1 and Ter2 sites. The major restriction sites in the area shown are as follows: RI, EcoRI; H, HindIII; Hp, HpaI; and Pv, PvuII. Bottom panel shows a magnified view of the features between the EcoRI and HpaI sites. REB1, the transcription terminator site that binds to Reb1p; ABF1, binding site for a transcription factor. IR and poly-T, inverted repeat and poly(dT) region that are involved in replication termination and HOT1 activity. N35, mutation in IR affects HOT1 activity and replication termination at Ter2. C20 and C26 affect HOT1 activity and replication termination at Ter1. II). When the blank plasmid (without FOB1) was present in the cell, there was very infrequent loss of ADE5 resulting in mostly solid red, non-sectoring colonies (Fig. 3B, panel I). We mutagenized the FOB1 sequence in vitro by PCR and introduced the mutagenized pool of plasmid DNA into the indicator host strain, and we looked for solid red colonies that potentially contained nonfunctional fob1 mutants. By specific site-directed mutagenesis, we also isolated mutant forms that did or did not knock out its recombinogenic activity. In this way we identified one mutation identified as L104S that generated solid red nonsectoring colonies (Fig. 2B, panel III). In contrast, a second mutation that was generated by mutating the GKT sequence in FOB1 to a GAT sequence (K168A) yielded a mutant form that did not lose the sectoring activity (Fig. 3B, panel IV). Although the GKT sequences are located in the putative Walker box indicative of a potential ATP-binding site, the FOB1 sequence did not have a typical Walker box (34). The K168A and a third mutation C246R (also known as HRM1-1 (21) that generated nonsectoring red colonies) are presented here as controls. A critical point in this context was to consider whether the mutants described above affected an isolated property of Fob1p (e.g. failure to terminate replication) or were trivial mutations that caused global misfolding of the mutant form of the protein.
We reasoned that global misfolding should affect almost all or many of the biochemical properties of the mutant form such as its ability to be translocated into the nucleolus (Fob1p is a nucleolar protein; Refs. 23 and 35), ability to oligomerize, and the ability to interact with other proteins. We therefore proceeded systematically to analyze the various properties of the mutant forms of Fob1p and the wild type control as described below.
Interaction of Fob1p with Itself and with Another Protein YDR026C-Fob1p is known to interact with itself (36) and with a protein encoded by a yeast gene called YDR026C that has no known function (37). We performed systematic two-hybrid analyses of the wild type and the mutant forms described above to investigate both homologous and heterologous protein-protein interactions. First, we proceeded to examine FOB1-FOB1 interaction of the wild type and the various mutant forms. During the course of our work we had noticed that intact wt FOB interacted less vigorously with itself as indicated by slower growth on the indicator plates. In contrast, a C-terminally truncated FOB1 (retaining the amino acid residues showing the two-dimensional gel pattern of replication intermediates of pBB3-H that was cut with SspI. The Y Ter1 and Y Ter2 are the spots generated by arrest of forks initiated at the ARS4 site and moving counterclockwise on the circular DNA at the Ter1 and Ter2 sites, respectively. X Ter1 and X Ter2 are X-shaped molecules that accumulate because of meeting of the forks stalled at the Ter sites with the forks moving clockwise. FIG. 3. HOT1 activity of wild type FOB1 gene and its mutants as revealed by the colony color sectoring assay. A, schematic diagram showing HOT1 activity. A duplication of leu2 mutants flanking an ADE5 gene and pBR322 sequences is present in chromosome III of the yeast strain used in the sectoring experiment. The 570-bp HOT1 element has been inserted in one of the copies of leu2. The strain that is ade2, ade5 and has the ADE5 gene (inserted in chromosome III) is red when it is non-recombinogenic (in the ⌬fob1 background). In the presence of a wild type FOB1 gene that acts at the HOT1 site, the ADE5 gene is lost from the cell by recombination and makes the cell white. Thus the strain with wild type FOB1 gives red colonies with white sectors. B, phenotype of wild type and mutant fob1 gene transformed into the ⌬fob1H strain. I, red colony produced by transformation of the vector pGADh. II, sectoring produced by transformation of plasmid containing wild type FOB1 gene. III, red colony produced by transformation of plasmid containing L104S; and IV, sectoring produced by transformation of plasmid containing K168A.
1-430) showed more vigorous self-interaction either between two truncated forms or between the truncated form and the intact protein. We also observed that the truncated form was functional in the sectoring assay described above and in replication termination (not shown). We used the truncated version of FOB1 and YDR026C as baits in the vector pGBT9 (with TRP marker) to look for protein-protein interaction. The intact FOB1 gene and the mutants L104S, K168A, and C246R were all cloned into pGAD424 (LEU2 marker). The results of twohybrid assay on SD/Trp Ϫ Leu Ϫ and SD/Trp Ϫ Leu Ϫ Ade Ϫ plates are shown in Fig. 4. Five colonies from each pair of plasmids were patched on each indicator plates. All cultures showed good growth on control plates (Fig. 4, A and C). In the FOB1-FOB1 oligomerization assay, only FOB1 and L104S showed good growth on SD/Trp Ϫ Leu Ϫ Ade Ϫ plates, whereas the vector, K168A, and C246R did not show any growth (Fig. 4, A and B). Similar results were obtained in SD/Trp Ϫ Leu Ϫ His Ϫ plates and in ␤-galactosidase assay in solution (not shown). Thus, we conclude from our experiments that both the wt and the L104S mutant form show self-interaction (oligomerization), whereas the K168A and the C246R mutant forms were completely defective in this interaction. We proceeded to analyze the interaction of wt and the various mutant forms of FOB1 with YDR026C. Initially, we had identified the YDR026C gene in a monohybrid assay that used tandem copies of yeast Ter1 sites cloned at the upstream activation sequence. A library of yeast DNA fused in-frame with the activation domain of Gal4 was transformed into the indicator strain to elicit activation of a His or a ␤-galactosidase reporter gene. In this way, YDR026C was identified as a potential DNA-binding protein that appeared to bind to the Ter region of yeast rDNA. We and others (37) had observed that YDR026C interacted with FOB1 in a two-hybrid assays. Here we have analyzed the interaction of wt and the mutant forms of FOB1 with YDR026C as an indicator of possible global misfolding caused by the mutations in Fob1p. In the two-hybrid analyses, wt FOB1, L104S, and K168A all showed interaction with YDR026C, whereas the blank vector and C246R showed no such interaction on SD/Trp Ϫ Leu Ϫ Ade Ϫ plates (Fig. 4, C and D). Similar results were also obtained on SD/Trp Ϫ Leu Ϫ His Ϫ indicator plates and confirmed by ␤-galactosidase assay. The results of two-hybrid analysis showed that L104S and K168A mutant proteins interacted with YDR026C, whereas the C246R mutant had lost that ability. The YDR026C locus encodes a Myb-like protein that in a monohybrid assay showed binding to rDNA. 2 Because YDR026C showed interaction with Fob1p, we wished to determine whether this interaction is needed for fork arrest. We performed a knockout of the YDR026C locus and performed two-dimensional gel assays to monitor possible fork arrest in the YDR026C-deleted strain. The results (Fig. 4, E and F) showed that both the wt and the isogenic YDR026C knockout strains were fully capable of fork arrest at the Ter sites. The results and other data discussed below showed that whereas Fob1p-Ter DNA interaction was definitely needed for fork arrest, none of the confirmed proteinprotein interactions with FOB1 mentioned were needed for replication termination.
The wt Fob1p and L104S and K168A Proteins Were Localized in the Nucleolus-Fluorescence microscopy of Fob1p fused to GFP and expressed in yeast showed that Fob1p is a nucleolar protein (23,35). We analyzed the intracellular localization of the wt and the mutant forms of FOB1 as another possible indicator of the folded state of the protein. We constructed N-terminal fusion of FOB1, L104S, and K168A with GFP under a Gal promoter (29). Cells with a modified pGADh plasmid, which expressed Fob1p-GFP or mutant fob1p-GFP, were grown in rich medium (yeast extract ϩ peptone) containing raffinose, galactose, and hygromycin. The cells were put on a microscope slide and photographed in phase contrast, in filtered UV light for DAPI staining or in GFP fluorescence. Fig. 5 shows the comparison of fluorescence studies among wt, L104S, and K168A. Merger of the DAPI and GFP fluorescence suggests that all of them are localized in the nucleus, and the GFP staining is further delimited to a part of the nucleus (e.g. see the merged pattern for wt in Fig. 5). The localized pattern of the GFP fluorescence within the nucleus along with published data (23,35) suggest that Fob1p was localized to the nucleolus. The results showed that the L104S and the K164A mutations did not abolish the transport and localization of the mutant forms of the protein into the nucleus. The protein-protein interaction data and the nuclear/nucleolar localization data, taken together, strongly suggested that the L104S mutation did not cause a global misfolding of the mutant form of the protein. interaction. E and F, two-dimensional gel analysis of wild type and ⌬YDR026C strains. Both wild type and ⌬YDR026C strains show similar termination spots. sectoring assay described earlier had shown that the L104S mutation abolished HOT1 activity. Is the loss of HOT1 activity due to failure to arrest replication forks? In order to answer this question definitively, we proceeded to perform two-dimensional Brewer-Fangman gel assays (32) of replication fork passage through the NTS1 region of rDNA. Host cells containing a deletion of the FOB1 locus were transformed with a plasmid pGADh that supplied either the wt or a mutant form of fob1, and the cells were grown to mid-log phase in rich medium containing hygromycin, harvested, and lysed. Replication intermediates were prepared by extracting whole cell DNA. The DNA was digested with BglII and resolved in neutral-neutral, Brewer-Fangman two-dimensional gels. The gels were blotted onto Nytran membranes and probed with a 32 P-labeled probe specific to the Ter region. The blank plasmid (without FOB1 insert) present in the ⌬fob1 host cells, as expected, did not generate a Ter spot. As expected, the wt FOB1 also elicited fork arrest as indicated by the presence of the Ter spot(s) just before the inflection point of the Y arc (Fig. 6). The L104S mutant failed to promote termination of replication as evidenced by the lack of a Ter spot (Fig. 6). The K168A protein on the other hand caused replication fork arrest (Fig. 6). Thus, the loss of fork arrest by the L104S mutation correlated with loss of HOT1mediated recombination. However, this defect was unlikely to be caused by a global misfolding of the mutant form of the protein because, as described above, other properties like FOB1-FOB1 interaction, FOB1-YDR026C interaction, or nuclear localization were unaffected by the mutation. The K168A mutant form retained replication fork arrest, HOT1 activity, and nuclear localization but had lost FOB1-FOB1 interaction properties and is therefore unlikely to have produced a misfolded protein. It should be noted that although K168A had lost Fob1p-Fob1p interaction, it was still able to arrest replication forks. The C246R mutant was negative on the basis of all of the criteria described above and was probably misfolded in the cell.
Purification of Fob1p Protein-We first constructed a GST-FOB1 fusion and attempted to express and purify this protein from E. coli cells. Although the expression was successful and the protein soluble, it was excessively prone to rapid proteolysis, and the protein failed to bind to Ter DNA with any degree of sequence specificity (not shown). An attempt to test the binding of Fob1p to the Ter sites by monohybrid analysis was marginally positive (not shown). These initial results had raised the possibility that Fob1p might bind to the Ter sites via interaction with a DNA binding adapter protein. However, subsequent expression and purification of the protein from yeast dispelled the notion of a possible involvement of a DNA binding, Fob1-interacting adapter protein. We expressed Fob1p as a TAP-FOB1 fusion in yeast. The TAP tag included an IgG-binding site followed by a tobacco etch virus protease cleavage site that in turn was followed by a calmodulin-binding peptide (28) fused in-frame to the reading frame of FOB1. The fusion protein was expressed under the control of an inducible Gal1 promoter. We have determined that the fusion protein was just as effective as the wt protein in promoting colony sectoring in the HOT1 assay and fork arrest in vivo as shown by two-dimensional gel analysis (data not shown). We have also expressed a fusion protein with both the GST and the TAP affinity tags under the Gal1 promoter. Expression of the fusion proteins was confirmed by Western blotting (data not shown). The protocol for purification of the GST-TAP-Fob1p protein is described in detail under "Experimental Procedures." At various steps of purification, including the final step, the protein fractions were assayed by Western blotting to confirm the identity of the protein. DNA binding assays were performed as described later to monitor the biological activity of the protein during purification. A Coomassie Blue-stained SDS-polyacrylamide gel shows the degree of purity of the protein (Fig. 7, lanes  1 and 2). The protein contains traces of a proteolytic fragment of Fob1p in addition to the intact protein.
Fob1p Specifically Bound to the Two Ter Sites-To test if Fob1p protein binds to Ter DNA by itself or in complex with other protein(s), we first tested the partially purified protein for DNA binding. We developed a new method for analyzing protein-DNA interaction as described below. The method is schematically shown in Fig. 8A. The cells expressing the TAPtagged Fob1p were lysed as described above, and the lysate, after removal of insoluble material and ribosomes by ultracentrifugation, was bound to calmodulin affinity beads in the presence of 2 mM CaCl 2 . The beads were thoroughly washed with several volumes of calmodulin binding buffer (buffer C; see "Experimental Procedures"). Then the beads were incubated for 15 min at 22°C with the specific DNA probe that consisted of an ϳ370-bp 32 P-labeled fragment containing the two Ter sites and nonspecific fragments generated by end labeling pUC18 DNA that was digested with HaeIII. The beads were washed, and the TAP-Fob1p protein-DNA complex was eluted with 2 mM EGTA as described above. The eluted samples were resolved in a non-denaturing polyacrylamide gel (Fig. 8B). The HaeIII fragments of pUC 18 showed no detectable binding under the experimental conditions to the beads containing Fob1p (Fig. 8B, lanes pUC-I and pUC-B). In contrast, when Ter DNA alone (Fig. 8B, lane Ter1,2 I) or Ter DNA mixed with the pUC18 fragments (Fig. 8B, lane pUCϩTer1,2 I) was loaded on the column, only the Ter fragment bound specifically to the protein on the column (arrow, Fig. 8B, lanes Ter1,2 B and  pUCϩTer1,2 B). In this experiment the eluted DNA was mixed with only DNA loading dye without SDS. In addition to a majority of free Ter DNA, another band characteristic of the Fob1p-DNA complex (asterisk, Fig. 8B) was also resolved. In order to make sure that the sequence-specific binding was caused by Fob1p and not by a Fob1p-accessory protein complex, we repeated the binding experiments with highly purified Fob1p and obtained identical results (not shown). Fob1p bound separately to both the Ter1 and Ter2 sequences, and the binding was physiologically significant. In order to further narrow down the binding sequences and to evaluate the physiological significance of the binding, we made use of several mutants at the Ter1 and Ter2 sites that selectively knock out Ter1 or Ter2 activity (7). The C26 mutant (and C20) that has a two-base substitution at the Ter1 site is known to inactivate Ter1 but not Ter2 and conversely the N35 mutation at the inverted repeats knocks out Ter2 but not Ter1 (7). We wished to determine whether a 40-bp (or 42-bp) DNA (oligonucleotide) that contains a functional Ter1 site specifically bound to Fob1p and if the C26 mutations (which includes C20 mutation) abolished the DNA binding activity. Sequences of 40-base oligonucleotides are shown in Fig. 9B. As shown in Fig. 9C, when the 42-bp oligonucleotide pair containing the wild type Ter1 was loaded (Fig.  9, TerI lane I) on the beads it bound to Fob1p protein (Fig. 9,  TerI, lane B). In contrast, when the double-stranded labeled oligonucleotide containing the C26 mutation (DNA containing the same Ter1 sequence with only 2-bp change; Fig. 9B) was loaded (Fig. 9, C26 lane I; the minor band below the major band is un-annealed labeled DNA), there was no binding of the DNA to the Fob1p protein Fig. 9, C26 B). Similarly, when a 40-bp DNA containing the M4M5 substitution (Fig. 9B) was loaded (Fig. 9, M4M5, lane I) on the column, there was no binding to the Fob1p protein (Fig. 9C, M4,5 B). We have carried out the oligonucleotide binding experiments with highly purified protein ( Fig. 7) with identical results (Fig. 9D). The Ter2 site includes three sequence segments: a poly(dT-dA) region, an inverted repeat (IR), and an inverted repeat-associated sequence (IRAS) as shown (Fig. 9A). Substitution mutations in the IRAS are known to inactivate Ter2 activity (7). The mutation N35 ( Fig. 1 and Fig. 10A) is known to knock out completely replication termination at Ter2 (7). Huang and Keil (20) have shown that the mutations C26 and N35 also reduce the HOT1 activity. We also wished to investigate possible binding of Fob1p to the inverted repeat site and the adjacent (IRAS). Oligonucleotides containing either the wild type IR or the N35 mutation were end-labeled and used to monitor Fob1p binding. When the 42-bp-long, inverted repeat oligonucleotide pair (IR) was loaded (Fig. 10B, IR, lane I) on the calmodulin beads containing Fob1p, it bound to the beads as shown in Fig. 10B, IR, lane B. However, when the (42 bp) DNA containing the N35 mutation (Fig. 10A) was similarly loaded (Fig. 10B, N35, lane I), the binding was drastically reduced (Fig. 10B, N35, lane B). We also designed 32-bp oligonucleotide pairs with mutations in the inverted repeat (IRm; Fig 10A), and when this DNA was FIG. 8. Technique that revealed sequence-specific interaction of Fob1p with the Ter1,2 sequences. A, diagrammatic representation of the experimental scheme. B, TAP-tagged Fob1p protein was bound to calmodulin beads in the presence of Ca 2ϩ , washed to remove adventitiously bound proteins, and incubated with 32 P-labeled pUC18 HaeIII fragments (used as a negative control), a labeled ϳ370-bp-long fragment containing both Ter1 and Ter2 DNA fragment or the mixture of pUC18 HaeIII and Ter fragments (shown as I in case of pUC, Ter1,2, and pUCϩTer1,2). The beads were washed to remove unbound DNA, and the affinity-bound protein-DNA complex was eluted with EGTA containing buffer. The input (I) and bound (B) DNA was fractionated in a non-denaturing 4% polyacrylamide gel and autoradiographed. The Ter fragment (shown by arrow) bound specifically to the Fob1p immobilized on the calmodulin beads and eluted with EGTA. The extra band shown by the asterisk is probably the protein-DNA complex between Fob1p and Ter DNA (no SDS was used to disrupt DNA-protein complex). loaded (Fig. 10B, IRm, lane I) onto the Fob1p beads, the binding was completely abolished (Fig. 10B, IRm, lane B). Finally, we designed a complementary pair of 30-bp oligonucleotides that contain the IRAS and tested its binding to Fob1p. The DNA was loaded onto the calmodulin beads containing Fob1p (Fig. 10B, IRAS I), and results show that Fob1p protein did not bind to the IRAS (Fig. 10B, IRAS, lane B). It would be interesting to investigate whether Fob1p nucleates at the IR sequence and then contacts the IRAS.
The L104S Mutation Significantly Reduced Binding of the Mutant Form of Fob1p to Ter DNA-Because the L104S mutant of Fob1p is defective in replication fork arrest, we wanted to test if it is manifested as a defect in DNA binding. We expressed TAP-tagged L104S and the wt proteins as described above. Equal amounts of wt and L104S cells were harvested, and cell extracts were prepared as described above and processed as shown schematically in Fig. 8A. Western blots of equal amounts of the wt and the mutant form bound to the calmodulin beads were performed (Fig. 11A) to make sure that both types of proteins were present in cell extracts and that they were soluble. Labeled TerI DNA was incubated with both types of immobilized proteins on the beads in the presence of CaCl 2 and eluted with EGTA containing buffer. Whereas wild type Fob1p bound to the TerI DNA, the mutant L104S showed a significant reduction in binding to the Ter1 probe (Fig. 11B). From these results we conclude that the L104S mutation abolished replication termination in vivo by greatly reducing the interaction of the mutant form of the protein with Ter1 DNA. DISCUSSION The mechanism of action of Fob1p protein of S. cerevisiae is of considerable interest not only because of its role in replica- FIG. 11. The L104S mutation that abolished in vivo replication termination also reduces Fob1p-Ter interaction. A, Western blots of the equal amounts of proteins from the wt and the L104S cells bound to calmodulin beads; B, autoradiogram showing results of 32 P-labeled Ter1 DNA binding with the wt and L104S proteins immobilized on calmodulin-agarose beads. The input and eluted fractions were resolved in a non-denaturing 8% polyacrylamide gel and autoradiographed and also quantified in a PhosphorImager. Note that the Leu-104 protein shows very low levels of binding to the DNA probe in comparison with the wt protein.
FIG. 12. A model of the Ter-Fob1p complex. The model hypothesizes that after interaction of Fob1p to the Ter sites, the DNA-protein complex contacts the putative replicative helicase of yeast (MCM proteins?). This contact need not be very specific. One orientation of the DNA-bound Fob1p (in analogy with bacterial replication termini) would make proper contact with the helicase and block its progression. In the other orientation, this contact would not occur, and the helicase would progress past the Ter sites, thus generating polarity of fork arrest.
FIG. 9. Mutations at the Ter1 site that abolish replication termination also eliminate or significantly reduce sequence-specific interaction of the Ter sites with Fob1p. A, features of DNA at the replication fork blockage sites. The Ter1 sequence is shown as a black rectangle. Ter2 sequence may encompass the poly(dT-dA) region, the inverted repeat, and the sequence known as RFB2 which we call here inverted repeat associated sequence (IRAS). B, the 40-bp sequence that encompasses the Ter1 region. The two blocks of 10-bp substitution (M4 and M5) are shown below the wild type sequence that is underlined. The C26 mutation has a C-T and T-C change. tion termination but also in cellular aging (23). In this paper, we show for the first time that Fob1p is a Ter-specific DNAbinding protein and that this binding is critical for replication fork arrest at the two Ter sites present in each NTS1 sequence of the rDNA array. Two lines of evidence support the abovementioned conclusion. First, a mutation L104S did not cause misfolding of the protein as revealed by retention of many of the biological activities of the mutant form of the protein but concomitantly abolished fork arrest and specific binding to the Ter1, Ter2 sequences. Second, several mutations in the sequences of Ter1 and Ter2 that separately knocked out the ability of Fob1p to terminate replication at the two sites abolished binding of the protein to the mutated sites. The demonstration that highly purified Fob1p specifically binds to the two Ter sites along with the results of fob1 mutants stated above provide compelling evidence that Ter DNA-Fob1p interaction is a critical step in replication termination at the Ter sites.
It should be noted that whereas Ter1 forms a single block of sequences that binds specifically to Fob1p, Ter2 consists of two essential sequence segments called IR and IRAS out of which only IR binds to Fob1p. However, mutations in both IR and IRAS abolish fork arrest (7). It is likely that IR forms a nucleation site for Fob1p binding that subsequently promotes binding of the same protein to IRAS. Experimental verification of this idea is in progress. Mutations in the poly(dT-dA) region of Ter2 do not appear to be essential for fork arrest but could play a minor accessory role in the efficiency of fork arrest (7).
Other clues to the mechanism of fork arrest could perhaps be surmised from a consideration of the mechanism of replication termination in bacteria. We have shown that both the interaction of Tus protein with the Ter sites of E. coli and interaction between DnaB helicase and the Tus protein are needed to cause polar fork arrest (5). Keeping this in mind, we present a hypothetical model of fork arrest in Fig. 12. The model suggests that Fob1p, bound to the Ter sites, could contact the hypothetical replicative helicase of yeast. At this time the identity of the replicative helicase that drives the fork in the region of rDNA is not known, although the hexameric MCM complex of yeast could turn out to be the replicative helicase (38 -41).
The MCM subunits have been expressed in baculovirus and in E. coli and have been purified (38,39). Recent work shows that a MCM sub-complex of S. cerevisiae has helicase activity (50). Similar results have been published for a MCM subcomplex of S. pombe (42). In addition to the MCM complex, two other helicases, namely the Pif1 and Rrm3, affect fork movement and/or replication termination in the rDNA region (43). It would be interesting to investigate whether the Fob1p-Ter complex arrests any of these helicases. Purification of Fob1p as described in this work should facilitate in vitro experiments designed to identify the replisomal target of the terminator protein. Could Fob1p arrest fork just by tight binding to Ter sites? In principle, a tight binding could explain fork arrest, but it would be more difficult to explain polarized fork arrest. In the future, it would be interesting to determine whether Fob1p physically contacts the MCM proteins should the latter turned out to be the replicative helicase.
Although the Ter1 and Ter2 sequences both bind to Fob1p, the two sequences have no detectable homology, thus raising the interesting question as to whether Fob1p has a single DNA binding domain that recognizes both sequences or whether there are two different domains each one recognizing one of the Ter sequences. Competition binding experiments suggest that Fob1p probably has a single DNA binding domain or a common domain that interacts with both Ter1 and Ter2. 3 Cross-linking experiments along with mass spectroscopies are in progress to map the amino acid to nucleotide contacts between Fob1p and the Ter sequences, which should yield results that could illuminate the question of the DNA binding domain(s) of Fob1p.
Two confirmed protein-protein interactions involving Fob1p have been discovered to date, none of which seem to be involved in fork arrest. First, Fob1p interacts with itself and thus is an oligomeric protein. However, oligomerization is apparently not required to terminate replication as exemplified by the K168A mutant form that is capable of effecting fork arrest but has lost the ability to oligomerize as determined in a two-hybrid assay. Second, Fob1p interacts, as shown in this work and by others (37), with YDR026C, which encodes a Myb-like protein. But the data presented here showed that the interaction is not involved in fork arrest.
A model building and homology search have suggested that Fob1p has similarity to the integrase sequence of retroviruses (44). In this context it is interesting to note that a plasmid (pBB3NTS; Ref. 7) with Ter sites can integrate into the chromosomal rDNA in the presence of a functional Fob1p and in the absence of a functional SIR2 gene (45), which encodes a histone deacetylase (46 -49). In the presence of wt SIR2, such integration was not observed (45). We have accumulated a large number of point mutations of fob1 including some at the region having similarity to the integrase sequence. We plan to investigate whether these mutants block the integration event and to determine whether the integration event is dependent on Fob1p-mediated fork arrest.
It should be pointed out that our work presents a novel technique for the detection of sequence-specific DNA-protein interaction using a calmodulin-binding peptide affinity tag and a calmodulin affinity column. This technique is highly sensitive and has very high signal to noise ratio and should be generally useful.
In conclusion, purification Fob1p and the demonstration that it specifically interacts with the Ter1 and Ter2 sequences provide critical evidence with regard to the mechanism of replication termination.