Both a Unique Motif at the C Terminus and an N-Terminal HEAT Repeat Contribute to G-Quadruplex Binding and Origin Regulation by the Rif1 Protein

Rif1 is a key factor for spatiotemporal regulation of DNA replication. Rif1 suppresses origin firing in the mid-late replication domains by generating replication-suppressive chromatin architecture and by recruiting a protein phosphatase.

S patiotemporal regulation of DNA replication permits ordered and coordinated duplication of the eukaryotic genomes (1). In higher eukaryotes, this regulation is associated with the higher-order chromosome organization in nuclei, which may be determined in early G 1 at TDP (timing decision point) (2). The mid-late replication domains appear to be sequestered from early replicating segments both temporally and spatially, so that they will not undergo premature early replication (3).
To identify potential new regulators of replication timing, we searched for bypass mutants for hsk1, the gene for the fission yeast homologue of budding yeast Cdc7 kinase, known to play crucial roles in initiation of replication (4,5). We had previously found mrc1 mutants can weakly bypass the Hsk1 function and also had shown that replication timing is altered in mrc1 mutants (6,7). The screening led us to identify rif1 as a strong bypass suppressor of the hsk1 null mutation (8). Furthermore, the origin firing pattern was dramatically altered in rif1Δ cells. Following this discovery, we and others reported that Rif1 is also a major regulator of genome-wide replication timing domain structures in mammalian cells and other species as well (9)(10)(11). We proposed that Rif1 regulates replication timing through generating chromatin architecture near the nuclear periphery that is suppressive for initiation (12). Circularized chromosome conformation capture (4C) analyses indicated that Rif1 defines and restricts the inter-actions between replication-timing domains, supporting the role of Rif1 as an organizer of nuclear architecture (13).
Rif1, originally identified as a telomere binding factor in budding yeast (14,15), has been shown to regulate transcriptional silencing and telomere length in concert with Rap1 (14,16,17). Rif1 protects telomeres through the CST (Cdc13-Stn1-Tel1) complex (18,19) and contributes to a telomeric anticheckpoint activity (20)(21)(22). Fission yeast Rif1, through interaction with Taz1, regulates telomere length and telomere position effect (TPE) (15). rif1Δ restores viability of taz1Δ at 25°C, although the molecular mechanism of this suppression is not clear (23). Roles of Rif1 in correct telomere length regulation have been reported in other yeasts as well (24,25). In contrast to that in yeasts, a direct role of Rif1 in telomere regulation is unlikely in mammalian cells, since it is not colocalized at telomeres and no obvious telomere defects were observed in Rif1 knockout cells (26,27). However, Rif1 indirectly regulates telomeres through transcriptional silencing of Zscan4, a factor promoting recombination-mediated telomere elongation in embryonic stem cells (28).
In human cells, Rif1 was found to accumulate at double-stranded DNA breaks (DSBs) in an ATM-and 53BP1-dependent manner (26,27). Rif1 also may participate in the intra-S-phase checkpoint to slow down DNA synthesis in response to DNA damage and also contributes to replication stress survival (27,29). More recently, Rif1, in conjunction with 53BP1, was shown to facilitate nonhomologous end joining (NHEJ) repair of DSBs by inhibiting 5= to 3= DNA end resection required for homologous recombination (HR)-mediated repair (30)(31)(32)(33)(34). Yeast Rif1, on the other hand, was shown to promote DSB resection, facilitating HR repair of DSBs (35). In Xenopus egg extracts, Rif1 is required for chromatin accumulation of several key checkpoint proteins (TopBP1, ATR, and the MRN complex) (36). Rif1 was also reported to function in resolution of nontelomeric chromosomal entanglement in M phase in fission yeast (37).
Only limited information is available as to how the structure of Rif1 and its biochemical functions are related to its diverse cellular functions. One of the conserved features of Rif1 is the presence of HEAT (Huntingtin, elongation factor 3, A subunit of protein phosphatase 2A, and TOR) or armadillo-type helical repeats (26). The N-terminal domain of Rif1 is predicted to have 14 to 21 tandem HEAT-like repeats (29), although the numbers and lengths of HEAT-like repeats depend on the species. Within the HEAT repeats, highly conserved Rif1-core segments have been identified. In fission yeast Rif1, an ϳ400-amino-acid (aa) segment (aa~100 to~500) is the most conserved among various HEAT repeats from many proteins, including Importin. Another conserved motif is the docking motif for type I protein phosphatase (PP1), consisting of SILK-RVXF sequences in multicellular organisms (38). In yeast, a PP1 binding motif can be found as an RVXF-SILK arrangement near the N terminus of the protein. The PP1-interacting motif was shown to be required for origin suppression activity both in yeasts and human (39)(40)(41)(42)(43).
It was reported that a C-terminal domain of Rif1 can bind preferentially to branched DNAs, including cruciform and forked DNA (29). It was also reported that a short segment of the C-terminal 60 amino acids (residues 1857 to 1916) of Saccharomyces cerevisiae Rif1 (ScRif1) forms a tetramer (44). We have reported that fission yeast Rif1 binds to specific segments on the chromosome (Rif1BS, for Rif1 binding site), whose sequences are characterized by the presence of multiple G-tracts, and that this binding is mediated by its specific recognition of G-quadruplex (G4) structure (45). The purified Rif1 protein binds to G4 with higher affinity than to cruciform DNA. We also purified the mouse Rif1 protein and showed that it also selectively and preferentially bound to G4 (46).
In order to elucidate how the functional domains on Rif1 protein are coordinated in its cellular activities, we genetically and biochemically characterized the truncated polypeptides of fission yeast Rif1 and also isolated loss-of-function mutants. Our results indicate the importance of both N-terminal and C-terminal domains for origin suppression and telomere regulation functions of Rif1 and requirement of their coordination for its in vivo chromatin binding and biological functions. Our results also identify critical residues for in vivo chromatin binding and origin suppression/telomere regulation. On the basis of biochemical properties of Rif1 and its truncation polypeptides, we will speculate on how it may contribute to the formation of replication-inhibitory chromatin architecture.

RESULTS
Both N-terminal and C-terminal segments of fission yeast Rif1 are required for its origin suppression activity. Rif1 is evolutionarily conserved, and the N-terminal segment is composed of many HEAT-and armadillo-type repeats (29), which are helical folds that form extended curved proteins or RNA interface surfaces. The N-terminal segment [Rif1(aa110-471)] is particularly well conserved and regarded as a Rif1 N-terminal domain. Rif1 has a PP1-interacting motif near its N terminus (Fig. 1A) (38). However, other structural and functional domains are not known. We previously reported that disruption of rif1 can suppress hsk1Δ, suggesting that loss of rif1 can bypass the requirement of Hsk1 function for growth (8). We have generated various truncation forms of Rif1 and examined if they can bypass Hsk1 function. We used an hsk1-89 mutant which does not grow at 30°C but can grow at 30°C under the rif1Δ background. The C-terminal 435-amino-acid truncation [Rif1(aa1-965)] as well as the N-terminal 442-amino-acid truncation [Rif1(aa443-1400)] resulted in efficient suppression of temperature sensitivity of hsk1-89 (Fig. 1B). Since the efficiency of suppression correlates with the loss of its activity to suppress origin firing, the results indicate both N-terminal and C-terminal segments of Rif1 are important for origin suppression. The truncated polypeptides were expressed at a similar level, except for Rif1(aa443-1400), which is unstable (Fig. 1C). The C-terminal 140-amino-acid truncation (aa 1 to 1260) resulted in partial suppression (Fig. 1B), indicating that the mutant still retains some activity to suppress origins. Consistent with this prediction, bromodeoxyuridine (BrdU) immunoprecipitation in the presence of hydroxyurea (HU) showed that origin firing at late origins was partially deregulated in rif1Δ1261-1400 [Rif1(aa1-1260)] (Fig. 1D). The chromatin immunoprecipitation-quantitative PCR (ChIP-qPCR) analyses indicated that Rif1(aa1-965) and Rif1(aa443-1400) completely lost their ability to bind to Rif1BS (Fig. 1E), whereas Rif1(aa1-1260) binds to the Rif1BS on the arm with affinity much lower than that of the full-length protein. The binding to telomere was almost completely lost in all truncated polypeptides (Fig. 1E). Both Rif1(aa151-1400) and Rif1(aa1-1260) failed to correct the telomere deficiency of rif1Δ cells, whereas the full-length Rif1 restored the normal telomere (Fig. 1F, lanes 6 to 8). This indicates that both N-terminal and C-terminal segments of Rif1 are required for its chromatin binding, for origin suppression as well as for telomere regulation.
Both N-terminal and C-terminal polypeptides of Rif1 bind to G4. We expressed various portions of Rif1, purified them ( Fig. 2A), and examined their G4 binding activity. All of the polypeptides containing either an N-terminal or C-terminal segment (aa 1 to 444, aa 1 to 1023, aa 472 to 1400, aa 1091 to 1400, and aa 1129 to 1400) bind to G4 DNA ( Fig. 2; see also Fig. S1 in the supplemental material).
Both an N-terminal polypeptide [Rif1(aa1-1023)] and a C-terminal polypeptide [Rif1(aa1091-1400)] bound to G4 with apparent dissociation constants (K d ) of 13.4 and 11.8 nM, respectively (Fig. S1). A shorter N-terminal polypeptide [Rif1(aa1-444)] also bound with a K d of 15.2 nM (Fig. S1). In contrast, the full-length Rif1 bound to G4 DNA with a K d of 0.3 to ϳ0.8 nM (Fig. S1), roughly consistent with a previous report (45). Thus, both C-terminal and N-terminal polypeptides bind to G4 with affinity lower than that of the full-length polypeptide. We noted that the Rif1(aa472-1400) polypeptide containing both the C-terminal DNA binding domain and a part of the HEAT repeats bound to G4 at a significantly higher level than other N-terminal or C-terminal polypeptides, with an apparent K d of 0.6 nM (Fig. S1C). Portions of conserved HEAT motifs present on this polypeptide appear to increase the affinity to G4.
DNA binding activity was previously reported for a N-terminal domain of S. cerevisiae Rif1 protein as well (44). Both N-terminal polypeptides [Rif1(aa1-1023) and Rif1(aa1-444)] generate large aggregates at the top of the gel and a small amount of fast- migrating shifted bands in gel shift assays ( Fig. 2C and E). The complexes that entered the gel appear to be somewhat smeared, suggesting that they are not very stable during electrophoresis. In contrast, the polypeptides containing the C-terminal segment generate discernible shifted bands with less large aggregates at the top of the gel (Fig. 2D). These observations suggest that the N-terminal and C-terminal polypeptides generate G4-protein complexes with distinct assembly.
With all the polypeptides, binding was competed most effectively by cold G4 DNA (T 6 G 24 ) itself (Fig. 3, lanes 9, 10, 21, 22, 33, 34, 45, and 46; Fig. S2 shows additional details) but not by other structured DNAs, including cruciform, forked DNA, and non-G4 DNA (T 6 [GA] 12 ). This was most clearly observed with the C-terminal polypeptides [Rif1(aa1091-1400) and Rif1(aa1129-1400)]. However, competition was weak with the N-terminal polypeptide [Rif1(aa1-1023)] under the same conditions (Fig. 3, lanes 21 and 22). In particular, the high-molecular-weight aggregates at the wells were not significantly reduced by addition of competitor G4 DNA (Fig. 3, lane 22). This could be due to the nature of the N-terminal polypeptides. In a native gel, they mostly run at the top of the gel, indicative of large aggregates (data not shown). In gel shift assays with polyacrylamide gel, the polypeptide aggregates may trap the added competitor DNA at the well. In pulldown assays, both Rif1(aa1-1023) and Rif1(aa1-444) polypeptides exhibited specificity to G4 DNA compared to that of non-G4 DNA (Fig. S3, lanes 4, 5, 9, and 10). The expected N-terminal polypeptides are enriched after pulldown with the G4 oligonucleotide, indicating the absence of other major DNA binders in contaminating polypeptides (Fig. S4, lanes 2 and 7). These results suggest that the dual G4 binding by the C-terminal and N-terminal segments of Rif1 support its G4 recognition/binding, and that the N-terminal HEAT/armadillo repeats facilitate the generation of high-molecularweight complexes through their self-associating activity.
Amino acid residues required for G4 binding and oligomerization of Rif1. Alignment of the C-terminal segment of Rif1 revealed the presence of conserved amino acid sequences, including ENSKKRQFSSLL at aa 1321 to 1332 and RLQRAILSR at aa 1390 to 1398 (Fig. 4A). We generated alanine substitutions at the conserved amino acids in the above-described two segments (underlined residues; mut1 and mut2, respectively) on the C-terminal polypeptide (aa 1129 to 1400). The purified mutant proteins were examined for G4 binding. Although mut1 polypeptide showed DNA binding activity identical to or greater than that shown by the wild-type polypeptide, mut2 showed a Ͼ10-fold reduction of DNA binding activity in gel shift assays (Fig. 4B, lanes 7 to 9 and 15 to 17; Fig. S5 provides additional data).
Since we previously showed that Rif1 forms oligomers (46; R. Fukatsu, K. Moriyama, and H. Masai, unpublished data), we then examined if the C-terminal polypeptide forms multimers and if the mutation affects the multimerization activity of the protein. We analyzed the purified proteins on a native protein gel. The C-terminal 272-aa (aa 1129 to 1400, 30 kDa [predicted molecular weight]), 229-aa (aa 1172 to 1400, 25 kDa), 148-aa (aa 1253 to 1400, 16 kDa), and 91-aa (aa 1310 to 1400, 10 kDa) polypeptides migrated are indicated by double-arrowed bars. The yellow segment represents conserved PP1 binding sites. (B) Tenfold serial dilutions of exponentially growing cells, as indicated, were spotted onto YES agar and incubated for 5 days at 25 or 30°C. (C, upper) Cell extracts were prepared from exponentially growing Flag-tagged Rif1, Rif1(aa1-1260), Rif1(aa1-965), Rif1(aa1-442), and Rif1(aa443-1400) cells, and the expression levels of truncated Rif1 polypeptides were estimated by Western blotting using anti-Flag antibody. (Lower) Anti-␣-tubulin antibody (loading control). (D) Cells expressing the indicated Rif1 truncation polypeptide were arrested at M phase by nda3-KM311 at 20°C and released into the cell cycle in the presence of 25 mM HU and 200 g/ml BrdU for 60 min at 30°C. The genomic DNA was extracted, and BrdU-substituted DNA was immunoprecipitated. IP efficiency (relative to the input genomic DNA) was assessed by quantitative real-time PCR. ars2004 is an early-firing origin, and Ori I:2663kb, Ori II:2580kb, Ori II:3380kb, and Ori II:4255kb are late/dormant origins that normally do not fire in the presence of HU. Non-Ori II:4282kb is a nonorigin sequence (negative control). Independent experiments were conducted three times. (E) Cells expressing the indicated Rif1 truncation polypeptide were arrested at M phase by nda3-KM311 mutation and released into the cell cycle at 30°C. The cells were fixed at 15 min after release. Rif1 chromatin immunoprecipitation was conducted, and the chromatin binding of Rif1 was assessed by quantitative real-time PCR at the sites indicated. I:2663kb and II:4255kb are known Rif1 binding sites, and II:4282kb is a negative-control site. Telomereassociated sequence is derived from the sequences located at the end of chromosomes 1 and 2. Independent experiments were conducted three times. (F) Wild-type (YM71), rif1Δ (FY14160), taz1Δ (FY14161), and rap1Δ (FY14171) cells were grown in YES. rif1Δ or rif1Δ cells expressing full-length or truncated Rif1 were grown in Edinburgh minimal medium without thiamine for 12 h. Cells were collected and genomic DNA was isolated by using a blood and tissue kit (Qiagen). Genomic DNA, digested by ApaI, was separated on 1% agarose gel, and then the telomere was detected by Southern hybridization using a probe containing the TAS1 sequence. at ϳ480 kDa, ϳ400 kDa, ϳ240 kDa, and 160 kDa, respectively, on a native gel (Fig. 5A,  lanes 6, 8, 9, and 10), suggesting that the C-terminal polypeptides form multimers. The molecular sizes predict the multimers potentially are composed of as many as 16 subunits, although it is difficult to accurately predict the numbers of subunits in each multimer, since the mobility on PAGE is affected by the shape of the molecule as well.
The results indicate that the C-terminal 91 amino acids are capable of multimerization, although they do not bind to G4 ( Fig. 5B; Fig. S6 provides additional data). Interestingly, the aa 1129 to 1400 mut2 polypeptide migrated at ϳ120 kDa (Fig. 5A, lane 7), suggesting that the mutant forms an oligomer with smaller numbers of subunits. In gel filtration, the wild-type aa 1129 to 1400 polypeptide migrated at around 500 to ϳ600 kDa, whereas the mut2 aa 1129 to 1400 polypeptide migrated at ϳ150 kDa (  S6), consistent with the results of the native gel, showing the importance of the C-terminal conserved sequences for oligomerization. These results indicate that the very C-terminal amino acid residues of Rif1 play a crucial role in G4 DNA binding as well as in oligomerization.
Further analyses of C-terminally truncated Rif1 polypeptides indicated that the C-terminal 229 amino acids (1172 to 1400) are sufficient for G4 binding, whereas the C-terminal 148-amino-acid polypeptide (aa 1253 to 1400) did not bind to DNA ( Fig. 5B; Fig. S6 provides additional data), indicating that the 81 amino acids (aa 1172 to 1252) are essential for G4 binding and that the ability to form a large multimer is not sufficient for G4 binding.

FIG 4
Mutations in the C-terminal conserved residues lead to reduced G4 binding. (A) The C-terminal amino acid sequences of Rif1 protein from three Schizosaccharomyces yeasts, S. cryophilus OY26, S. octosporus yFS286, and S. pombe, are aligned. Red, green, and blue letters indicate the identical, conserved, and related amino acid residues, respectively. Numbers below the sequences indicate the amino acid numbers of the S. pombe Rif1. The extents of the four C-terminal polypeptides analyzed in this study are shown, with numbers indicating the residues at their N termini. Two clusters of well-conserved amino acid residues are underlined. These segments were mutated by changing the amino acids shown in boldface letters to alanine (mut1 and mut2). (B) The labeled T 6 G 24 DNA (0.25 pmol), heat denatured and renatured in 50 mM KCl and 40% PEG 200, was incubated with various amounts of full-length protein (lanes 2 to 5), the wild type (aa 1129 to 1400; lanes 7 to 9), mut1 (aa 1129 to 1400; lanes 11 to 13), or mut2 (aa 1129 to 1400; lanes 15 to 17), and samples were analyzed by 8% PAGE (1ϫ TBE, 50 mM KCl, and 40% PEG 200). The graph shows quantification of the binding, conducted as described in the legend to Fig. S1. Mutations of the conserved sequences near the C terminus of Rif1 (mut2) reduce the G4 binding activity in gel shift assays (estimated K d , ϳ250 nM). The result presented is a representative from two independent experiments, and the graph represents the average values. Another set of results is shown in Fig. S5.
Complexes between Rif1 polypeptides and G4. The presence of dual DNA binding domains suggests a single molecule of Rif1 binds to more than one molecule of G4. In order to assess more precisely the number of G4 molecules bound by a single molecule of Rif1, we conducted gel shift assays in the presence of excess amounts of the G4 oligonucleotide. The amount of the bound substrate DNA was determined by measuring the radioactivity of the bound (complex) and unbound (free) DNA bands.
In order to get better separation of bound and unbound G4, we used a different gel condition (10% polyethylene glycol 200 [PEG 200] instead of 40%). Under this condition, the bound G4 migrates near the well, giving a better separation from the unbound G4. Another feature of this new gel system is that it can separate oligomers of the T 6 G 24 DNA, known to from a parallel-type G4. 32 P-labeled T 6 G 24 DNA (2.5 pmol) was used as a substrate (final concentration of 250 nM). Under this gel condition, the T 6 G 24 DNA appeared as a clear ladder, indicative of the presence of dimer, trimer, tetramer, etc. In titration of the C-terminal polypeptides (aa 1091 to 1400 and aa 1129 to 1400), the larger T 6 G 24 DNAs were bound at the lowest concentration of the polypeptide, and the smaller G4 elements were bound only at a higher concentration (Fig. 6A), as was predicted from the results for other gels (Fig. 2). These results indicate the preference of the Rif1 C-terminal G4 binding domain for the polymers of G4.
A molecule of the above-described C-terminal polypeptides bound close to two molecules of the G4 DNA, on average, at a low protein concentration (12 nM). This is  Fig. S1. aa 1172 to 1400 can bind to G4 but not aa 1253 to 1400 or aa 1310 to 1400. In this experiment, the polypeptides derived from C-terminal segments were expressed and purified from E. coli. The result presented is a representative of two independent experiments, and the graph represents the average values. Another set of results is shown in Fig. S6. probably due to its preferential binding to the oligomers of G4. At a high concentration (ϳ300 nM; almost equimolecular to the G4 DNA), the number of G4 molecules per molecule of C-terminal polypeptide dropped to 0.3 to 0.5. The results suggest that the stoichiometry of G4 per monomer of the C-terminal polypeptide is low; the polypeptide likely binds to a single molecule of G4. In contrast, the preparation containing the  Fig. S1, except that 2.5 pmol of labeled T 6 G 24 DNA was used as substrate DNA. The samples were analyzed by 8% PAGE (29:1) in 1ϫ TBE, 50 mM KCl, and 10% PEG 200. After an autoradiograph was taken, dried gel pieces containing the free and complex DNA bands were excised from each lane, and the radioactivity was measured in a scintillation counter. The radioactivity in the complex band was divided by the sum of radioactivity of the free and complex DNA, and the amount of template DNA bound with the polypeptide was estimated. On the assumption that all of the polypeptides are involved in the binding, the number of the G4 molecules per molecule of the polypeptide was calculated for each binding (insert tables). (B) Full-length and truncated polypeptides, as shown, were used for gel shift assays, as described for panel A. The number of G4 molecules per molecule of each polypeptide was calculated at different concentrations of peptide, as described for panel A, and the values are presented in the table. Under this condition, Ͼ90% of the polypeptides added in the reactions are present in the bound complexes (confirmed by Western blotting).
full-length polypeptide bound to the G4 substrate with stoichiometry of Ͼ15 G4 molecules at 1 nM protein and Ͼ8 G4 molecules per molecule of the polypeptide even at 4 nM protein (Fig. 6B). Interestingly, the N-terminal polypeptide (aa 1 to 1023) also showed stoichiometry of ϳ7 G4 (at 2 nM) and Ͼ3 G4 (at 8 nM) molecules. In contrast, the number of G4 molecules per molecule of the C-terminal polypeptide did not exceed 4 even at 1 nM and was less than 2 at 8 nM (Fig. 6B). By Western blot analyses of the same binding samples, we have confirmed that Ͼ90% of the polypeptides added are involved in binding to G4 at a concentration of 8 nM (data not shown). These results suggest a role of the HEAT domain in coordinating the binding of Rif1 to multiple G4 molecules in conjunction with the C-terminal G4 binding domain rather than the possibility that the two G4 binding domains on Rif1 independently bind to G4.
Identification of point mutant Rif1 capable of bypassing Hsk1 function. In order to identify novel motifs involved in Rif1's origin suppression function, we screened mutants of Rif1 which could rescue the growth of hsk1-89 at 30°C (Fig. S8 depicts the strategy). Six mutants were identified, and mutation sites were determined (Fig. S9). Generally, the mutants contain multiple amino acid substitutions, and we tried to identify the responsible mutations by constructing mutants harboring each singleamino-acid substitution. We have identified R236H and L848S, each of which alone could suppress hsk1-89. L848S suppressed hsk1-89 to an extent similar to that by rif1Δ, while R236H did so slightly weakly ( Fig. 7A and Fig. S9). The single L848S mutation suppressed hsk1-89 as efficiently as the L848S T865A double mutant, whereas T865A, which coexisted with L848S in mut4, did not restore the growth of hsk1-89 at 30°C (data not shown); thus, the L848S T865A double mutant can be regarded as being identical to the L848S single mutant. We next examined the chromatin binding of these mutant Rif1 proteins at two Rif1BS (Rif1BS I:2663 and Rif1BS II:4255 ) and a telomere (NUS70_TEL). R236H bound to both Rif1BS and to the telomere as efficiently as or more efficiently than the wild type. On the other hand, L848S did not bind to either of the two Rif1BS or to the telomere. T865A bound to chromatin as efficiently as the wild type, consistent with its no-effect phenotype (Fig. 7B). mut5 (S707P D731G I778V V1265A) bound to the two Rif1BS less efficiently but bound to telomere as efficiently as the wild type (Fig. 7B). BrdU incorporation was deregulated in the L848S mutant at four late-firing origins, Ori I:2663 , Ori II:2580 , Ori II:3380 (ars727), and Ori II:4255 , as vigorously as in rif1Δ (Fig. 7C). In the R236H mutant, DNA synthesis at Ori I:2663 and Ori II:2580 , but not at Ori II:3380 or Ori II:4255 , was weakly deregulated (Fig. 7C). The elongated telomere in rif1Δ was corrected by the wild-type as well as by the R236H mutant Rif1 protein. However, the L848S mutant could not restore normal telomere. The mut5 mutant partially restored telomere length (Fig. 7D). Thus, the L848S mutant, incapable of chromatin binding, has lost the functions for both origin suppression and telomere regulation, whereas R236H, capable of binding to chromatin, appears to be defective more specifically in origin regulation at some selected late-firing origins. The results establish that chromatin binding of Rif1 is essential for its capability for origin firing, and they indicate a role of additional interfaces (in addition to N-and C-terminal domains) on Rif1 protein in its chromatin binding and that not only chromatin binding but also other unknown functions of Rif1 are required for its origin suppression activity.
Fission yeast Rif1 C-terminal segment contains both G4 DNA binding and multimerization activities. Rif1 binding sites (Rif1BSs) on the chromosome, as determined by ChIP sequencing (ChIP-seq), contains conserved sequence motifs (Rif1CSs) composed of 5 to 6 G tracts. Mutagenesis analyses indicated functional significance of (B) Cells carrying the indicated rif1 mutation (tagged with His 6 -Flag 10 at the C terminus) were arrested at M phase by nda3-KM311 at 20°C and released into the cell cycle at 30°C. The cells were fixed 15 min after release. Rif1 chromatin immunoprecipitation was conducted, and the chromatin binding of Rif1 was assessed by quantitative real-time PCR at the sites indicated. I:2663kb and II:4255kb are known Rif1 binding sites, and II:4282kb is a negative-control site. NUS70_TEL is located at a telomere. (C) Cells carrying the indicated rif1 mutation were arrested at M phase by nda3-KM311 at 20°C and released into the cell cycle in the presence of 25 mM HU and 200 g/ml BrdU for 60 min at 30°C. The genomic DNA was extracted, and BrdU-substituted DNA was immunoprecipitated. The IP efficiency (relative to the input genomic DNA) was assessed by quantitative real-time PCR. The origins examined are the same as those described for Fig. 1D. In panels B and C, independent experiments were conducted three times. (D) The indicated cells were collected and genomic DNA was isolated by using a blood and tissue kit (Qiagen). Genomic DNA, digested by EcoRI, was separated on 1% agarose gel, and then telomere was detected by Southern hybridization using a probe containing the TAS1 sequence. these G tracts for Rif1 binding and for G4 formation in vitro (45). Purified Rif1 binds selectively to G4 in vitro, suggesting direct recognition of G4 by Rif1. Gel filtration and glycerol gradient analyses indicate that Rif1 forms tetramers, octamers, or even larger oligomers (our unpublished data).
Analyses of polypeptides derived from the C-terminal segments on a native gel clearly indicated the formation of a distinct oligomer (Fig. 5A), which was estimated to be a multimer containing 16 subunits at maximum. The C-terminal 91 amino acids were sufficient for generation of a multimer. We have mapped the G4 DNA binding domain within the C-terminal 229-amino-acid polypeptide, and a further 81-amino-acid deletion led to loss of DNA binding. Furthermore, amino acid substitutions of the conserved sequences near the C terminus (mut2) resulted in reduced DNA binding and partial destabilization of oligomer formation. mut2 generated an oligomer of reduced size, probably a tetramer or dimer. The analyses on gel filtration also supported this conclusion (see Fig. S7 in the supplemental material). Thus, the C-terminal conserved sequences are required for both high-affinity G4 binding and stable formation of a large oligomer ( Fig. 4 and 5), suggesting that G4 binding is functionally linked to multimer formation. Since the shapes of the polypeptides affect migration on a native gel or in gel filtration, we cannot rule out the possibility that these C-terminal polypeptides are tetramers or octamers (with mut2 aa 1129 to 1400 being a monomer or a dimer).
It has been reported that the C-terminal segments of Rif1 from budding yeast and mammalian cells contain DNA binding activity and can oligomerize. aa 2170 to 2246 of human Rif1 was reported to bind to fork and Holliday junction (HJ) DNA (29). The C-terminal 60 aa (CTD) from budding yeast Rif1 protein forms a tetramer and binds to Rap1 (44). More recently, CRII (aa 2226 to 2340) from mouse Rif1 was reported to constitute a structure-specific DNA binding domain (48). Although this domain has similarity to the tetramerization domain of budding yeast Rif1, it did not show clear oligomerization activity.
The N-terminal segment of Rif1 also binds to DNA in a G4-specific manner. We discovered that the N-terminal HEAT motif segments also contain DNA binding activity ( Fig. 2 and Fig. S1). In gel shift assays on G4, the N-terminal polypeptides tend to generate high-molecular-weight complexes (trapped at the top of the gel) as well as a small amount of lower-molecular-weight complexes that appear as broad bands (Fig. 2). Indeed, the majority of the N-terminal polypeptides migrate as high-molecularweight complexes (aggregates at the well) on a native gel (data not shown). In pulldown assays, the N-terminal HEAT motif polypeptides also showed specific binding to G4 (Fig. S3). Recently, structures of the N-terminal domain of S. cerevisiae Rif1 (ScRif1-NTD) were solved (47). The ScRif1-NTD adopts a unique shepherd's crook shape, the hook of which is involved in capturing DNA. ScRif1-NTD forms a dimer and was proposed to spread on DNA. Similar mechanisms may operate on high-molecularweight complex formation by fission yeast Rif1 N-terminal polypeptides on G4. ScRif1-NTD was shown to be essential for telomere length regulation and for the telomeric anticheckpoint function (47).
In contrast to the N-terminal polypeptides, the C-terminal polypeptides generate distinct shifted bands on G4. These results reveal aspects on the molecular nature of the N-terminal and C-terminal polypeptides, which may interact with DNA in distinct manners. The affinity of the full-length Rif1 to G4 is at least severalfold higher than that of the C-terminal or N-terminal fragment (Fig. S1), suggesting that the N-terminal HEAT and C-terminal G4 binding domains (delineated to 444 and 229 aa, respectively) are coordinately involved in high-affinity and stable binding of Rif1 to DNA. The dual G4 binding domain structure is also shared by mammalian Rif1 protein (46).
Dissection of functional domains of Rif1 protein in regulation of replication timing. In spite of aberrant spatiotemporal regulation of DNA replication, fission yeast rif1Δ cells are quite healthy and grow normally. They do not show sensitivity to replication stress or to DNA damage agents (data not shown). We have shown that the ability of rif1Δ to bypass Hsk1 (or to suppress temperature sensitivity of hsk1-89) reflects the loss of origin suppression activity; therefore, we utilized this characteristic of rif1Δ to measure the origin suppression activity of each mutant (Fig. 8).
Analyses of truncation mutants indicate that deletion of either the N-terminal 442 or C-terminal 435 amino acids results in complete suppression of hsk1-89, i.e., loss of origin suppression (Fig. 1B). Neither truncation mutant binds to Rif1BS in vivo (Fig. 1E), supporting the roles of both N-terminal and C-terminal segments in chromatin binding. The deletion of the C-terminal 140 amino acids, which is a part of the C-terminal G4 binding domain, led to partial suppression of hsk1-89 and caused firing at late-firing origins at the level 20% to 50% of that observed in rif1Δ. The mutant also exhibited Chromatin binding represents binding to Rif1BS on the arm as determined by ChIP-qPCR/ChIP-Seq. G4 binding represents in vitro binding of the purified protein to the G4 DNA. (B) Schematic drawing of the functional segments of fission yeast Rif1 protein determined in this study. Among the three conserved armadillo/HEAT repeat segments (shown by double-arrowed bars) identified by using the "SUPERFAMILY" search method, the longest armadillo/HEAT repeat (aa 114 to 514) largely overlaps the Rif1 core sequence, the segment showing the highest conservation from yeast to human. Two novel mutations of rif1 that impair origin suppression activity are shown. The conserved RLQRAILSR (aa 1390 to 1398) sequences, important for G4 binding and multimerization, are indicated. Deletion of either N-terminal (Δ1-442; aa 443 to 1400) or C-terminal (Δ966-1400; aa 1 to 965) sequence led to complete loss of origin suppression activity. Deletion of the C-terminal 140 amino acids (aa 1 to 1260) also led to partial but significant loss of origin suppression and chromatin binding. significantly reduced chromatin binding at Rif1BS (20% to 25% at the Rif1BS arm and less than 10% at telomeres) (Fig. 1E). The N-terminal 442 amino acids contain PP1 binding sites (KVNF-GILK; present at aa 40 to 67), known to be essential for origin suppression; thus, loss of origin suppression in Rif1(aa443-1400) is likely to be caused by its inability to recruit PP1. However, R236H mutation, located in the conserved HEAT repeat and not expected to affect PP1 recruitment, impairs origin suppression, supporting the role of the N-terminal HEAT repeat in origin suppression. The HEAT repeats, which are required for telomere and Rif1BS arm binding, as shown above, are required for localization of Rif1 at DSB in mammalian cells as well (32). Thus, diverse cellular roles of Rif1 depend on the conserved N-terminal HEAT repeats. Results described above indicate the functional significance of G4 binding mediated by both N-terminal (HEAT domain delineated to 444 aa) and C-terminal (delineated to 229 aa) segments of Rif1 in its origin suppression activity. The combined actions of the two domains may generate a nucleoprotein complex required for replication timing regulation.
How does Rif1 contribute to the formation of chromatin architecture that is suppressive for origin activation? Quantitative analyses of Rif1 binding to G4 suggest a single molecule of the C-terminal G4 binding polypeptide binds to a single molecule of G4 (Fig. 6). In contrast, a single molecule of the full-length Rif1 protein appears to interact with multiple molecules of G4 DNA. Furthermore, the N-terminal polypeptide containing HEAT repeats (aa 1 to 1023) is capable of binding to multiple molecules of G4, albeit with reduced stoichiometry compared to that of the full-length version. We speculate that each subunit of the full-length Rif1 protein interacts with multimerized G4 through its C-terminal G4 binding domain and the shepherd's crook-shaped N-terminal HEAT repeats, which may be interacting with those from the adjacent assembly of Rif1. The combined action of the two domains would lead to a stable complex between Rif1 and G4 multimers. Rif1 forms a multimer through its C-terminal domain, and each subunit would bind to the G4 multimers. The target G4s are derived from the Rif1BS on the genome, which we speculate would interact with each other to form G4 multimers. The chromatin fibers between the Rif1BS would be looped out to generate specific chromatin architecture that would be essential for proper replication timing regulation (Fig. 9).
Point mutants of Rif1 capable of bypassing Hsk1 suggest the presence of additional interfaces required for its origin suppression activity. We took advantage of the abovementioned rif1Δ genotype to isolate novel mutants. Selection of viable hsk1-89 cells at 30°C after mutagenesis of the rif1 gene led to identification of 6 mutants that can suppress the hsk1 mutation (Fig. S8). Mutants generally contained multiple amino acid changes. After separating each mutation, we came up with two mutations, L848S and R236H, each of which is responsible for Hsk1 bypass (Fig. 7). L848S does not bind to Rif1BS or telomere in vivo, whereas R236H retains full binding activity (Fig. 7B). L848S also is defective in telomere regulation, but R236H is mostly proficient in telomere regulation (Fig. 7D). PP1 recruitment may not be affected in either mutant, since PP1 binding sites are intact.
The L848S mutant lost origin suppression and telomere regulation activities, probably due to its inability to bind to chromatin. It is not clear how L848S mutation affects chromatin binding of Rif1. The purified L848S protein binds to G4 as efficiently and selectively as the wild-type protein in vitro (data not shown), suggesting that G4 binding alone is not sufficient for chromatin binding of Rif1 in vivo. Stable binding of Rif1 to chromatin requires not only G4 binding but also some other unknown factors that may be affected by the L848S mutation. The R236H mutation does not affect the in vivo chromatin binding but is partially defective in origin suppression. In vitro, the R236H mutation does not affect DNA binding activity of the full-length or of the N-terminal polypeptide (data not shown). Thus, chromatin binding is not sufficient for origin suppression, and some other factors that interact with R236 would be required. Alternatively, a part of the HEAT motif structure may be partially disrupted by the mutation, which is located in the middle of the most conserved sequences. Both L848 and R236H are within the segments that are conserved in fission yeast species (data not shown); thus, they may constitute unknown functional interfaces required for the functions of Rif1. Further biochemical and genetic analyses of the mutants would be required to clarify how Rif1 binds to chromatin and regulates chromatin architecture.

MATERIALS AND METHODS
Medium for Schizosaccharomyces pombe. YES medium, containing 0.5% yeast extract, 3% glucose, and 0.1 mg/ml each of adenine, uracil, leucine, lysine, and histidine, was used for cell culture, and YES plates were made by adding 2% agar to YES medium. G418 (0.2 mg/ml) was added to YES medium for selection of KanMX. For 5-fluoroorotic acid (5-FOA) selection, 0.1 mg/ml 5-FOA was added to medium containing 6.3 g/liter synthetic dextrose minimal medium (SD), 2% glucose, and 0.1 mg/ml each of adenine, uracil, and leucine. Yeast strains used in this study are listed in Table 1.
Genetic manipulations for construction of truncation mutants of Rif1. First, a plasmid containing 5=UTR-Rif1-His 6 -Flag 10 -kanMX-3=UTR sequences was constructed. With the inverse PCR on this plasmid and In-Fusion reaction (TaKaRa Bio Inc.), plasmids carrying 5=UTR-Rif1 (with a truncation)-His 6 -Flag 10 -kanMX-3=UTR sequences were generated. DNA fragments containing 5=UTR-Rif1 (with or without a truncation)-His 6 -Flag 10 -kanMX-3=UTR sequences were used to generate tagged wild-type rif1 or rif1 truncation mutants at the endogenous rif1 locus of the wild-type or hsk1-89 cells (see Fig. S8 in the supplemental material). In the presence of the N-terminal HEAT repeats, Rif1 may make additional interactions through the N-terminal domain, causing a single molecule of Rif1 to interact with multiple G4 molecules. Rif1 is drawn as a dimer for simplicity. (Center) Oligomer formation (octamer) of the C-terminal G4 binding domain generates the Rif1 bead structure. The portion indicated by a bracket represents a putative octameric Rif1 molecule. (Right) Oligomerized Rif1 (octamer) may interact with multiple chromatin fibers through G4 binding, generating chromatin loops that may constitute replication-inhibitory chromatin domains. Here, we speculate that G4 generated in the intergenic segments on the genome interact with each other, generating multimerized G4, which contributes to the formation of chromatin loops.
DNA binding assays were conducted twice or more with similar results, and representative data are shown in the main figures (additional sets of the data for gel shift assays presented in main figures are shown in Fig. S1, S2, S5, and S6).
Pulldown assays. Ten picomoles of biotinylated T 6 G 24 or T 6 (GA) 12 oligonucleotide and purified SpRif1 derivative polypeptides, as indicated, was mixed in 200 l of 40 mM HEPES-KOH (pH 7.6), 50 mM KCl, 1 mM EDTA, 10% glycerol, 1 mM dithiothreitol, and 0.01% Triton X-100 and incubated at room temperature for 30 min. After addition of 10 l of streptavidin Dynabeads (M-280), equilibrated in the buffer described above, biotinylated oligonucleotides were pulled down, washed with 500 l of binding buffer three times, resuspended in 20 l of 1 ϫ SDS-PAGE sample buffer, boiled at 96°C for 1 min, and analyzed by 5% to 20% gradient SDS-PAGE, followed by Western analyses with anti-Flag antibody.

SUPPLEMENTAL MATERIAL
Supplemental material for this article may be found at https://doi.org/10.1128/MCB .00364-18.