Pol κ: A DNA Polymerase Required for Sister Chromatid Cohesion

See allHide authors and affiliations

Science  04 Aug 2000:
Vol. 289, Issue 5480, pp. 774-779
DOI: 10.1126/science.289.5480.774


Establishment of cohesion between sister chromatids is coupled to replication fork passage through an unknown mechanism. Here we report that TRF4, an evolutionarily conserved gene necessary for chromosome segregation, encodes a DNA polymerase with β-polymerase–like properties. A double mutant in the redundant homologs, TRF4 and TRF5, is unable to complete S phase, whereas a trf4 single mutant completes a presumably defective S phase that results in a failure of cohesion between the replicated sister chromatids. This suggests that TRFs are a key link in the coordination between DNA replication and sister chromatid cohesion. Trf4 and Trf5 represent the fourth class of essential nuclear DNA polymerases (designated DNA polymerase kappa) inSaccharomyces cerevisiae and probably in all eukaryotes.

We identified the TRF4gene in a genetic screen for functions redundant with DNA topoisomerase I (1) and showed that, together with its close homolog,TRF5, the genes are essential for several events in DNA metabolism including chromosome segregation (2,3) and DNA damage repair (4). Recent analysis of the highly conserved TRF4 gene family has led to the conclusion that TRFs are members of the β-polymerase superfamily (5), which consists of proteins that catalyze a variety of nucleotidyltransferase reactions including DNA synthesis.

To test the prediction that TRFs possess nucleotidyltransferase activity (5), we purified Trf4 fused to a six-histidine tag from Escherichia coli to apparent homogeneity (6,7) and assayed recombinant protein for DNA polymerase activity. First, Trf4 was examined for the ability to extend a 5′ end-labeled oligo(dT) primer (16 mer) that was hybridized to a poly(dA) template [average size: 282 nucleotides (nt)] in the presence of deoxythymidine triphosphate (dTTP) and Mg2+(Fig. 1A). Trf4 is able to extend the primer in a distributive manner (extension of a single nucleotide followed by dissociation from primer/template), which is characteristic of β-DNA polymerases (8). In contrast, a mutant Trf4 protein missing the NH2-terminal 240 amino acids of the 584–amino acid protein, Trf4Δ240, was completely unable to polymerize nucleotides (Fig. 1A, lanes 5 to 7).

Figure 1

(A) DNA polymerase assays with an oligo(dT)/poly(dA) primer/template. Equal amounts of purified recombinant Trf4-6HIS or Trf4Δ240-6HIS were incubated with 100 nM oligo(dT)/poly(dA) [20:1 molar excess of poly(dA):oligo(dT)] for 5 min at 37°C in 50 mM tris-HCl (pH 8.0) (pH at 22°C), 10 mM MgCl2, 2 mM DTT, 20 mM NaCl, 20 mM KCl, 2.5% glycerol, bovine serum albumin (0.2 mg/ml), and 1 mM dTTP. No protein was added to lane 1. The concentration of both Trf4 and Trf4Δ240 is 150 nM in lanes 2 and 5, 15 nM in lanes 3 and 6, and 1.5 nM in lanes 4 and 7. (B) Cofractionation of Trf4 protein and DNA polymerase activity on a mono Q anion exchange column. DNA polymerase assays were performed as in (A). The top panel shows a Coomassie-stained SDS–polyacrylamide electrophoresis gel of the same column fractions. Immunoblots confirm the identify of the protein as Trf4. (C) Trf4 DNA polymerase activity is not sensitive to a neutralizing monoclonal antibody to DNA pol I. DNA polymerase assays were performed as in (A) in a total volume of 20 μl. Lanes 1 and 4 contain no added antibody, assays shown in lanes 2 and 5 contain 1 μg of DNA pol I neutralizing monoclonal antibody 7C5.4, and assays shown in lanes 3 and 6 contain 1 μg of DNA pol I nonneutralizing monoclonal antibody 15B2.3.

Fractions eluted from the final purification step, a mono Q anion exchange column, demonstrate cofractionation of Trf4 protein and DNA polymerase activity (Fig. 1B). The DNA polymerase activity observed is dependent on template, primer, and Mg2+. To ensure that the activity observed was not due to E. coli pol I contamination [the major polymerase activity in E. coli extracts (9)], we used a neutralizing monoclonal antibody to DNA pol I in the polymerase reactions (10). Figure 1C shows that incubation of the neutralizing antibody with DNA pol I inhibited its ability to extend the oligo(dT) primer, whereas a different monoclonal antibody to DNA pol I that does not neutralize the activity had no effect on DNA pol I activity (10). In contrast, Trf4 activity was unaffected by either monoclonal antibody (Fig. 1C, lanes 4 to 6). In addition, the size range of the Trf4 products was consistently observed to be greater than for DNA pol I (Fig. 1C).

Sensitivity of the Trf4 DNA polymerase to a variety of inhibitors used to classify DNA polymerases was determined (11,12). The Trf4 activity was resistant to aphidicolin (50 μg/ml) and to preincubation with the sulfhydryl-blocking agentN-ethylmaleimide (NEM; Fig. 2A). However, in the presence of equimolar amounts of dideoxythymidine triphosphate (ddTTP) and dTTP, the Trf4 activity was inhibited (Fig. 2A). Resistance to aphidicolin and NEM and sensitivity to dideoxynucleotide inhibition are properties of β-DNA polymerases (11, 13). Furthermore,E. coli DNA pol II and pol III are sensitive to NEM and pol II is sensitive to aphidicolin (9), excluding contamination by these proteins as the source of the observed activity.

Figure 2

(A) Sensitivity of Trf4 DNA polymerase to sulfhydryl-blocking agent NEM, aphidicolin, and dideoxynucleotide triphosphates (ddNTPs). DNA polymerase assays were performed as in Fig. 1A with the following modifications: lane 1, no modification; lane 2, aphidicolin (50 μg/ml) present during assay; lanes 3 and 4, preincubation in 1 mM NEM for 1 or 10 min, respectively, before assay; lanes 5 and 6, preincubation in 10 mM NEM for 1 or 10 min, respectively, before assay; lanes 7, 8, and 9, ddTTP:dTTP present during assay in molar ratios of 1:1, 5:1, and 10:1, respectively. (B) Trf4 DNA polymerase can use a complex DNA primer/template as substrate. A template oligonucleotide of 75 nt was hybridized to a complementary 35 mer (pair 1, lane 1), 36 mer (pair 2, lane 2), 37 mer (pair 3, lane 3), or 38 mer (pair 4, lane 4) labeled with 32P at its 5′ end with T4 polynucleotide kinase in vitro. Extension of the primer/template substrate (present at 10 nM) was monitored by electrophoresis in an 8% polyacrylamide gel containing 8 M urea, after addition of Trf4 (15 nM) for 5 min at 37°C under conditions described in Fig. 1A, except that reactions contained 10 mM each of the four deoxynucleotide triphosphates (dNTPs): dTTP, dATP, dGTP, and dCTP. (C) Trf4 DNA polymerase activity is template directed. Reaction conditions are as described in (A), except that only one of the four dNTPs was added to each individual reaction. Each reaction contains primer/template present at 10 nM. In panel 1, the substrate used was the primer/template described in (A) (35 mer/75 mer hybrid or pair 1). Panel 2 contains the reaction products derived from a 36 mer/75 mer primer/template (pair 2) in which the primer oligonucleotide is exactly 1 nt longer than the 35 mer used in panel 1. Panel 3 products are derived from a 37 mer/75 mer (pair 3) and panel 4 from a 38 mer/75 mer hybrid (pair 4).

Trf4 is also able to use a complex DNA template (Fig. 2B). The oligo/template pairs, which differ in length by 1 nt at the 3′ end, were hybridized separately to the same 75 mer template oligonucleotide after the oligo primer was 5′ end-labeled with 32P. Template/primer pairs were then incubated with Trf4 in the presence of all four dNTPs, and the products were analyzed on a urea-acrylamide gel. The polymerization reactions in each of these primer/template pairs were relatively processive, because by far the most abundant extension product is the approximate size of the template. This is in contrast to the distributive polymerization observed in the oligo(dT)/poly(dA) assay.

To determine whether Trf4 DNA polymerase activity is template directed, we hybridized the oligo primers used in Fig. 2B separately to the same 75 mer template oligo. The various primer/template junctions formed are illustrated in Fig. 2C. Extension of template/primer pair 1 by Trf4 is observed only in the presence of deoxycytidine triphosphate (dCTP), indicating that the polymerization reaction is not only template-dependent but also template-directed. Similarly, pair 2 is extended only in the presence of deoxyguanosine triphosphate (dGTP), pair 3 only with dTTP, and pair 4 only with deoxyadenosine triphosphate (dATP), demonstrating that all four deoxynucleotides are recognized by Trf4 and used as substrates for DNA synthesis in a template-directed manner.

Our previous genetic studies with TRF4 indicate that it is required for mitotic chromosome segregation (1–4). One link between DNA replication and chromosome segregation is the establishment of cohesion between sister chromatids, which occurs during DNA replication (14–16) and is crucial for accurate segregation at anaphase. Furthermore, if DNA replication occurs in the absence of the cohesin SCC1/MCD1, expression of SCC1/MCD1 after replication is insufficient to establish cohesion (14), suggesting a direct coupling of replication to cohesion. We have previously observed thatTRF4 displays both genetic and physical interactions with Smc1 (2), a protein required for sister chromatid cohesion (17, 18).

To determine whether TRF4 is required for sister chromatid cohesion, we monitored the state of cohesion in otherwise isogenicTRF4 + and trf4 mutant cells with two different assays. In the first assay, fluorescence in situ hybridization (FISH) of a biotin-labeled cosmid DNA probe located 225 kb from the centromere of chromosome XVI (“CEN distal” probe in Table 1) was used to examine sister chromatid cohesion (19). In yeast cells in which cohesion is effectively maintained, a single fluorescence signal is observed because of the close proximity of the sister chromatids (19). Cells were first arrested with the microtubule-depolymerizing drug nocodazole, which blocks the cell cycle in mitosis after the generation of sister chromatids, but before cohesion is dissolved during anaphase. Wild-type cells are known to maintain sister chromatid cohesion at the nocodazole block (19). FISH was then performed on over 50 nuclei from both wild-type and the trf4 mutant.

Table 1

Quantitation of cohesion defect in the trf4mutant. Cultures were grown to early logarithmic phase in YPD medium; half of the culture was arrested with α-factor and the other half with nocodazole. After 4 hours, cells were fixed and processed for FISH (28). Nuclei were hybridized with either a chromosome XVICEN-proximal probe (11 kb from CEN16) or aCEN-distal probe (225 kb from CEN16) as described (2). The number of FISH signals was counted for each nucleus and expressed as the percentage of nuclei containing two signals for each condition. A total of 50 to 100 nuclei were counted for each probe from two independent experiments.

View this table:

As expected, the wild-type haploid strain showed only one hybridization signal in the majority of nuclei (87%), indicating that sister cohesion was intact (Fig. 3A and Table 1). The 13% of wild-type nuclei that display two or more signals has also been observed by others (16, 19,20) and may represent partial disruption of cohesion by the fixation procedure. In contrast, the trf4 mutant cells showed two hybridization signals in 41% of nuclei, demonstrating that cohesion is frequently defective in the absence of TRF4function. Simultaneous 4′,6′-diamidino-2-phenylindole (DAPI) staining was used to confirm that the two fluorescence signals observed in the trf4 mutant were derived from a single nucleus (Fig. 3A). FISH assays of a known cohesion protein,MCD1/SCC1 (20), show a similar fraction (50 to 60%) of nuclei that fail to maintain cohesion under these conditions (20). The two hybridization signals that are observed intrf4 mutant nuclei are the result of sister cohesion failure and are not due to the presence of two copies of chromosome XVI (7).

Figure 3

(A) A trf4mutant is defective for sister chromatid cohesion. Wild-type andtrf4 mutant cells were arrested in mitosis with nocodazole, and two representative nuclei are shown after hybridization with a fluorescently labeled centromeric cosmid probe from chromosome XVI. The blue signal in the “DAPI” panel shows nuclear DNA, and the green signal in the “FISH” panel shows the positions of sister chromatids from chromosome XVI in wild-type and mutant nuclei. (B)TRF4/TRF5 function is required for maintenance of cohesion. Wild-type and trf4-ts trf5 double-mutant cells containing GFP-marked chromosome III were arrested in mitosis with nocodazole at 30°C and then shifted to 37°C for 3 hours. Cells were then examined by fluorescence microscopy for the status of sister chromatid cohesion. (C) A trf4-ts trf5 double mutant is unable to complete DNA replication. Wild-type and trf4-ts trf5 double-mutant cells were subjected to centrifugal elutriation to obtain a homogenous population of G1 daughter cells. Cells were released into YPD medium prewarmed to 37°C, and samples were withdrawn every 20 min to examine DNA content, budding index, and viability. (D) Atrf4-ts trf5 double mutant dies during an abortive S phase. Viability of wild-type and trf4-ts trf5 double-mutant cells was monitored each hour after G1 daughter cells were shifted to 37°C by dilution and plating at the permissive temperature on YPD medium.

To examine sister cohesion at other chromosomal regions, we used a 35-kb centromere-proximal cosmid probe that spans a region between 11 and 45 kb from the centromere on the left arm of chromosome XVI (“CEN proximal” probe in Table 1). A similar cohesion defect is observed in the trf4 mutant cells with this probe [Table 1 (7)], indicating that TRF4 is required for cohesion both near centromeres and on chromosome arms.

We also used a second assay to monitor the cohesion status of a different chromosome. In the “green fluorescence protein (GFP)–marked chromosome” assay (21), chromosomes are visualized with immunofluorescence microscopy by theGFP fluorescence signal generated when a GFP-lacrepressor fusion is expressed in a yeast strain bearing 256 tandem repeats of the lac operator integrated on one of the chromosomes (21). Of the over 100 nuclei examined for each data point in these experiments, two signals were present in only 2.5% of wild-type nuclei (Table 2). In contrast, the trf4 mutant showed two signals in 18% of cells [Table 2 (7)], seven times more frequently than in wild type. Furthermore, the cohesion defect in trf4is anaphase promoting complex–independent because it is still observed in a cdc16 mutant strain [Table 2(7)], demonstrating that failure of the spindle assembly checkpoint is not the reason for its cohesion defect. Cohesion occurs at multiple distinct sites along each chromosome, with a concentration of sites near the centromere (22). Thus, chromosomes with defective cohesion at some sites and not others are likely to segregate normally. Therefore, it may not be surprising that the trf4 single mutant remains viable. Consistent with this explanation, we observed that the mcd1-1 ts cohesion mutant also shows a clear cohesion defect even at the permissive temperature of 30°C (7). Thus, TRF4 is required for sister chromatid cohesion as determined by two independent assays, on both chromosomes XVI and III and at both centromere-proximal and -distal loci.

Table 2

Quantitation of cohesion defect in the trf4mutant. Strains containing the lacI-GFP fusion and 256 tandem repeats of the lacO integrated at the LEU2locus on chromosome III were grown in SC-his to induce the Lac-GFP repressor; cells were pelleted, resuspended in YPD with nocodazole, and incubated for an additional 4 hours. Cells were fixed and adhered to glass slides. The number of fluorescent signals (GFP) was counted for each nucleus and cell. At least 80 nuclei were counted for each strain from two independent experiments.

View this table:

To determine whether TRF function was required to maintain cohesion, we examined cohesion in the trf4-ts trf5double mutant. If TRF function were required to maintain cohesion, then cells arrested with nocodazole at a permissive temperature and then shifted to the nonpermissive temperature should show an increased cohesion defect. After nocodazole arrest at the permissive temperature of 30°C, the trf4-ts trf5double mutant was able to maintain cohesion as monitored by the GFP-marked chromosome assay. However, when mutant cells were shifted to 37°C for 3 hours, a marked increase in cohesion-defective cells was observed (Fig. 3B). This demonstrates that TRF function is indeed required to maintain cohesion.

The ability of cells lacking TRF function to duplicate the genome was then examined as follows. Synchronous populations of either wild-type or otherwise isogenic trf4-ts trf5double-mutant G1 daughter cells were obtained by centrifugal elutriation. Wild-type and mutant cells were then released into yeast extract, peptone, and dextrose (YPD) medium at 37°C, and DNA content, budding index, and cell viability were monitored during a single synchronous cell cycle. Figure 3C shows that wild-type cells begin DNA synthesis about 60 min after release and complete replication roughly 100 min after release, as expected. In contrast, the trf4-ts trf5 double-mutant cells show a marked delay in the G1/S transition with little evidence of DNA synthesis onset for 120 min (Fig. 3C). A similar G1/S transition delay has been observed in themcd1-1 mutant (20), and we also observed that thetrf4 mcd1-1 double mutant is completely inviable at any temperature (7). This G1/S delay is likely to represent a checkpoint-induced phenomenon because bud emergence is also delayed in both cases. Three hours after release, the majority of trf4-ts trf5 mutant cells contain between G1 and G2 DNA content (Fig. 3C), and viability has dropped 10-fold (Fig. 3D). Thus, cells lackingTRF function do not completely replicate the genome.

Evidence that other cohesins are involved in establishment of cohesion is based primarily on the observation of S phase–specific lethality (15, 16). During the single cell cycle described above, the trf4-ts trf5 mutant cells die rapidly and are unable to complete S phase (Fig. 3D). This is likely to reflect a requirement for TRF function in the establishment of sister chromatid cohesion. However, if chromatids are not fully duplicated, they are naturally held together by unreplicated regions, and, furthermore, as soon as a cell completes replication it is not possible to distinguish between a role in establishment and maintenance of cohesion. Thus, we envision that replication in the absence of onlyTRF4 is able to proceed because of TRF5 function but that this enfeebled replication fork fails to properly establish cohesion.

To determine whether DNA polymerase activity per se of Trf4 is required for proper cohesion, we made a site-directed mutation in the putative active site motif, DXD, that is known from the crystal structure of rat DNA polymerase β to coordinate Mg2+ and to be important for catalysis (13). The trf4-236 allele replaces both aspartate residues (Trf4 amino acids 236 and 238) with alanine. DNA polymerase assays of the recombinant Trf4-236 performed side by side with Trf4 show that the mutant protein's activity is greatly diminished in the oligo(dT)/poly(dA) assay (Fig. 4A). To examine the role of Trf4 polymerase activity in cohesion, we introduced the trf4-236mutation into the GFP-marked chromosome strain. After nocodazole arrest, the mutant shows a clear cohesion defect compared with an otherwise isogenic TRF4 + control (Fig. 4B). Residual polymerase activity present in Trf4-236 or a partial separation of the polymerase and cohesion functions may explain why the magnitude of the defect is not as great as that observed in thetrf4Δ strain.

Figure 4

(A) Trf4-236 has reduced DNA polymerase activity. Equal amounts of Trf4 or Trf4-236 were assayed for DNA polymerase activity as inFig. 1A. Lanes 1 and 2 contain 10 nM Trf4 and Trf4-236, respectively, lanes 3 and 4 contain 5 nM, lanes 5 and 6 contain 2.5 nM, and lanes 7 and 8 contain 1.25 nM. The Trf4 and Trf4-236 proteins were purified identically. (B)trf4-236 is partially cohesion defective. Cohesion assays in wild-type and trf4-236 mutant cells with the GFP-marked chromosome (21, 29). (C)trf4-236 is inviable in the absence ofTRF5. A trf4-236 trf5/pTRF5.URA3strain (CY1262) or an otherwise isogenic TRF4 +parent strain (CY1231) was grown nonselectively for theTRF5.URA3 plasmid to single colonies on YPD medium and examined for the presence of spontaneous segregants that had lost the plasmid with 5-FOA counterselection. The TRF4 +strain is able to remain viable in the absence of TRF5 as segregants can grow on 5-FOA medium (left panel), whereas thetrf4-236 strain cannot survive in the absence ofTRF5 because it is unable to grow on 5-FOA medium.

The question of whether Trf4 DNA polymerase activity is its overlapping essential function with TRF5 was then addressed as follows. The trf4-236 allele was integrated to replace the wild-typeTRF4 at the natural locus in strains containing either a wild-type or mutant TRF5 allele, but expressingTRF5 from a plasmid to ensure viability (7). The Trf4-236 protein was found to be expressed at a wild-type level by immunoblotting (7). The ability of the trf4-236 allele to maintain viability in the absence of TRF5 was then monitored by counterselection of the TRF5 plasmid with 5-fluoro-orotic acid (5-FOA). Thetrf4-236 allele is inviable in the absence ofTRF5 expression because it is unable to form colonies on 5-FOA medium (Fig. 4C). Thus, the Trf4 DNA polymerase activity is also required for the overlapping essential function of TRF4 andTRF5.

The common TRF4/5 function represents only the fourth essential nuclear DNA polymerase in yeast and probably in all eukaryotes. The previously observed lethality in trf4 top1double mutants (1) is likely a result of the combined loss of both the primary DNA replication swivel (TOP1) and another key replication factor (TRF4). Assays of purified human pol κ [hTRF4-1 gene product (4)] also show it to be a DNA polymerase (23). We propose calling the yeast and human Trf4 proteins DNA polymerase κ. Recently, an S. pombe TRF4/pol κ family member, cid1, was found to show genetic interactions with DNA polymerase δ and ɛ and involvement in an S phase checkpoint (24). Therefore, the TRF4/pol κ gene family may be involved in multiple aspects of chromosome metabolism.

Genetic studies in yeast have implicated the DNA polymerase processivity factor PCNA in the establishment of cohesion (16), and studies of meiotic DNA synthesis and cohesion also suggest a link between the two processes (25). Thus, it will be important to determine whether pol κ activity is affected by PCNA. The S. pombe eso1 gene, which is required to establish sister chromatid cohesion, consists of two domains, one that is related to the S. cerevisiae cohesion establishment gene CTF7/ECO1 and a second related to DNA polymerase η (26). An attractive model is that cohesion establishment factors require physical proximity to a DNA polymerase in order to execute their function. It has been suggested that repair polymerases such as pol η may travel with the replication fork and allow bypass of damaged regions that would otherwise cause fork collapse or create a terminal pause site (12). It may be that the establishment of cohesion is analogous in the sense that encounter of the replication fork with a cohesion site results in a switch in the DNA polymerase used to polymerize these regions. Although there have been major insights into the cohesion process recently (27), many questions remain regarding precisely how replication fork passage results in the establishment of effective cohesion. Our results indicate that a DNA polymerase, pol κ, provides an important link between replication and cohesion and should facilitate further elucidation of the mechanisms involved.

  • * Present address: Johns Hopkins University, Department of Molecular Biology and Genetics, 725 North Wolfe Street, 504 PCTB, Baltimore, MD 21205–2185, USA.

  • To whom correspondence should be addressed. E-mail: mfc3f{at}


View Abstract

Navigate This Article