Structure and Processivity of Two Forms of Saccharomyces cerevisiae DNA Polymerase δ*

Yeast DNA polymerase δ (Polδ) consists of three subunits encoded by the POL3, POL31, andPOL32 genes. Each of these genes was cloned under control of the galactose-inducible GAL1–10 promoter and overexpressed in various combinations. Overexpression of all three genes resulted in a 30-fold overproduction of Polδ, which was identical in enzymatic properties to Polδ isolated from a wild-type yeast strain. Whereas overproduction of POL3 together withPOL32 did not lead to an identifiable Pol3p·Pol32p complex, a chromatographically distinct and novel complex was identified upon overproduction of POL3 andPOL31. This two-subunit complex, designated Polδ*, is structurally and functionally analogous to mammalian Polδ. The properties of Polδ* and Polδ were compared. A gel filtration analysis showed that Polδ* is a heterodimer (Pol3p·Pol31p) and Polδ a dimer of a heterotrimer, (Pol3p·Pol31p·Pol32p)2. In the absence of proliferating cell nuclear antigen (PCNA), Polδ* showed a processivity of 2–3 on poly(dA)·oligo(dT) compared with 5–10 for Polδ. In the presence of PCNA, both enzymes were fully processive on this template. DNA replication by Polδ* on a natural DNA template was dependent on PCNA and on replication factor C. However, Polδ*-mediated DNA synthesis proceeded inefficiently and was characterized by frequent pause sites. Reconstitution of Polδ was achieved upon addition of Pol32p to Polδ*.

The subunit structure of eukaryotic DNA polymerase ␦ (Pol␦) 1 remains ambiguous (for a review, see Ref. 1). The most thoroughly characterized form of mammalian Pol␦ is that isolated from calf thymus, a two-subunit enzyme with a 125-kDa catalytic subunit and a 48-kDa accessory subunit (2). The accessory subunit is required for efficient stimulation of Pol␦ by the proliferating cell nuclear antigen (PCNA) (3,4). Likewise, mouse Pol␦ can be purified in two forms, the single catalytic subunit form which is not stimulated by PCNA and the twosubunit enzyme which is stimulated by PCNA (5). The subunit composition of the two-subunit enzymes is that of a heterodimer (2,5). In contrast, Pol␦ isolated from the two yeasts is more complex with the enzyme from Saccharomyces cerevisiae having three subunits and that from Sshizosaccharomyces pombe at least four, and perhaps five subunits (6,7).
The three subunits of S. cerevisiae Pol␦ have apparent sizes by SDS-PAGE of 125, 58, and 55 kDa and are encoded by the POL3, POL31, and POL32 genes, respectively (7). The 125-kDa catalytic subunit encoded by the POL3 (CDC2) gene is very highly conserved in all eukaryotes (1). The 58-kDa second subunit is encoded by the POL31 (HYS2, SDP5) gene (7). Mutations in this essential gene can cause sensitivity to the replication inhibitor hydroxyurea, for the hys2-1 allele, or suppress the temperature sensitivity of mutations in the catalytic subunit, for the sdp5-1 allele (8,9). Pol31p shows 23-28% sequence similarity to the 48-kDa subunit of human Pol␦ and to S. pombe Cdc1. The essential cdc1 ϩ gene encodes the second subunit of S. pombe Pol␦ (6). The 55-kDa subunit is encoded by the POL32 gene. Mutants deleted for POL32 are viable, but show both replication and repair defects (7). Although the sequence similarity between Pol32p and Cdc27, which is the essential third subunit of S. pombe Pol␦, is very low, other considerations including the presence of a PCNA-binding motif in both subunits denote these two as functional homologues (discussed in Ref. 7).
In the previous paper (7) we have described the cloning of the two small subunit genes of Pol␦ and their characterization.
Here we describe the overproduction in yeast of the threesubunit form of Pol␦, and of a two-subunit form, called Pol␦*, which is analogous to mammalian Pol␦.
The overproduction plasmids used in this study are based upon the pRS420 series plasmids into which the GAL1-10 upstream activating sequence (GAL1-10 upstream activating sequence) including the transcriptional start sites for the GAL1 and GAL10 genes, as a 678-nt BamHI-EcoRI fragment, was inserted into the corresponding plasmid polylinker sites, resulting into vectors pRS424-GAL (TRP1), pRS425-GAL (LEU2), and pRS426-GAL (URA3) (10). All vectors have in addition the yeast 2 M origin for high copy maintenance in yeast and the Bluescript SKII ϩ backbone for propagation in E. coli. The transcriptional start site of the GAL1 gene is 60 nt upstream of the BamHI cloning site and the transcriptional start site of the GAL10 gene is 10 nt upstream of the EcoRI cloning site. Both promoters are of similar strength. Coordinates are with reference to the translational start sites. pBL336 (TRP1 GAL1-POL3) has a 3.6-kb HgiAI (trimmed)-HindIII fragment (coordinates: Ϫ45 to 3543) cloned into the BamHI (filled)-HindIII sites of pRS424-GAL. pBL338 (LEU2 GAL1-POL31) has a 1.6-kb NcoI (filled)-ClaI (filled) fragment (coordinates: 2 to 1567) from pBL361 cloned into the SacII site of pRS425-GAL (7). pBL340 (URA3 GAL10-POL32) has a 1.7-kb HpaI-SalI fragment (coordinates: Ϫ20 to 1688) from pBL384 cloned into the EcoRI (filled)-SalI sites of pRS426-GAL (7).

Cell Growth
A single colony of a plasmid-containing strain from a selective SCGL plate was grown in an air shaker at 30°C in 100 ml of selective SCGL medium. SCGL medium contains per liter: 1.7 g of yeast nitrogen base without amino acids and ammonium sulfate, 5 g of ammonium sulfate, 30 ml of glycerol, 20 ml of lactic acid, 1 g of glucose, 20 g of agar for solid media, 20 mg each of adenine, uracil, histidine, tryptophan, proline, arginine, and methionine, 30 mg each of isoleucine, tyrosine, and lysine, 50 mg of phenylalanine, and 100 mg each of leucine, glutamic acid, aspartic acid, valine, threonine, and serine. Uracil, tryptophan, and/or leucine were omitted when appropriate to ensure the selective maintenance of plasmids. Prior to autoclaving, the pH of the media was adjusted to 5-6 with concentrated sodium hydroxide. After 2-3 days when the OD 660 had reached 0.8 -1, the culture was used to inoculate 1200 ml of SCGL media. After overnight growth, when the OD 660 was about 1, 1200 ml of YPGL were added. YPGL contains per liter: 10 g of yeast extract, 20 g of peptone, 30 ml of glycerol, 20 ml of lactic acid, 2 g of glucose, and 20 mg of adenine. Prior to autoclaving, the pH of the media was adjusted to 5-6 with concentrated sodium hydroxide. The culture was equally divided over two 4-liter flasks and grown at 30°C for 3 h. Solid galactose (2% final concentration) was then added to each flask and after 4 h of continuous shaking the cells were harvested.
Bead beating was carried out in a 350-ml chamber containing 175 ml of glass beads (0.4 -0.5 mm diameter) and 80 -100 g, wet weight, of cells, resuspended in an equal volume of 2 ϫ buffer A. The chamber was cooled in ice water and the beater turned on for 45 s, followed by a cooling period of 2 min, for a total beating time of 5 min. The lysate was poured in a cold graduated cylinder and the beads were washed with 50 ml of extraction buffer. An aliquot was centrifuged at 45,000 ϫ g for 20 min (cleared lysate, see Table I). The volume of the crude lysate was measured and 40 l of 10% Polymin P were added per ml of lysate. After 5 min of mixing, the lysate was spun for 40 min at 13,000 rpm in a GSA rotor. Solid ammonium sulfate (0.28 g/ml) was added to the supernatant and dissolved by stirring. The precipitate was collected at 13,000 rpm for 45 min. The pellet was resuspended in 20 ml of Buffer B. After dialysis against 2 ϫ 500 ml of buffer B for 8 h each, the dialysate was cleared by centrifugation at 18,000 rpm for 20 min.
The cleared ammonium sulfate fraction was loaded on a 20-ml phosphocellulose column, equilibrated in buffer B. The column was washed with 40 ml of B 25 and eluted with B 750 . The protein-containing fractions were combined and dialyzed for 2 ϫ 3 h against 150 ml each of buffer C until the conductivity of the dialysate was equal to that of C 25 .
The dialyzed fraction was injected onto a 8-ml MonoQ column, equilibrated in buffer C 25 , washed with 10 ml of C 25 , and eluted with a 120-ml linear gradient from C 25 to C 500 . Fractions of 2.5 ml were collected, Pol␦* eluted at ϳC 150 and Pol␦ at ϳC 200 .
Individual Mono Q fractions were diluted with 2 volumes of buffer D and injected onto a 1-ml MonoS column, equilibrated in buffer D 50 . The column was washed with 2 ml of D 50 and eluted with a 15-ml linear gradient from D 50 to D 500 . Pol␦* eluted at ϳD 250 and Pol␦ at ϳD 400 .
Samples of 200 l were injected onto a 20-ml Superose 6 column in 40 mM Hepes-NaOH, pH 7.5, 10% ethylene glycol, 1 mM EDTA, 0.02% Nonidet P-40, 0.2 M NaCl, 5 mM DTT, 5 mM NaHSO 3 , and 2 M each of leupeptin and pepstatin A. The column was run at 0°C at 0.2 ml/min. Fractions of 300 l were collected and analyzed by 10% SDS-PAGE and for DNA polymerase activity.

Enzyme Assays
DNA Polymerase-The DNA polymerase assay on activated DNA is described in the previous paper (7). Incubations were for 5 min at 37°C. The assays were stopped with 7 l of 95% formamide, 20 mM EDTA and electrophoresed on a 12% denaturing polyacrylamide gel.

RESULTS
Overexpression of the Subunits of Pol␦-An inducible system for the overexpression of the Pol␦ genes allows normal cell growth without possible deleterious effect on cells due to constitutive high levels of the Pol␦ subunits. The strain used in this study is the protease-deficient strain BJ2168. In this strain, the expression of genes placed under control of the GAL1-10 upstream activating sequence is appropriately induced by addition of galactose to the media, but the strain grows very poorly on galactose as sole carbon source. Satisfactory cell growth was obtained on media containing as carbon source 3% glycerol, 2% lactate, and a non-repressing concentration of glucose (0.1%). Galactose was added to this media to induce expression. The three Pol␦ genes were cloned under control of the bi-directional GAL1-10 upstream activating sequence as described under "Materials and Methods." Constitutive overproduction of Pol␦ is inhibitory to yeast cell growth as a strain carrying the three overexpression plasmids grew less well on galactose medium than on raffinose, a non-inducing carbon source (data not shown).
Cells containing one or more of the Pol␦ genes under GAL1-10 control were grown up to mid-logarithmic phase, induced with galactose, and crude extracts were made as described under "Materials and Methods." These extracts were assayed for DNA polymerase activity. However, because Pol␣ constitutes the major polymerase activity in yeast extracts, a moderate overproduction of Pol␦ would not be readily apparent when assayed in crude extracts. Therefore, Pol␣ activity was specifically inhibited with the dGTP analog BuPhdGTP. Under those conditions approximately one-half of the total DNA polymerase activity in crude extracts from a protease-deficient strain can be ascribed to Pol␦ (34).
Overexpression of the POL3 gene alone resulted in an over 100-fold increase in POL3 mRNA levels, a strong increase in Pol3p polypeptide levels, and an 8-fold increase in BuPhdGTPresistant DNA polymerase activity in cleared lysates (Table I, Fig. 1, data not shown). However, this polymerase activity was unstable and most of the increased activity was lost upon further fractionation by ammonium sulfate precipitation and phosphocellulose chromatography. A Western analysis showed that most of the Pol3p polypeptide precipitated during dialysis of the redissolved ammonium sulfate pellet, and some precipitated during dialysis after phosphocellulose chromatography (data not shown). These data indicate that the catalytic polypeptide of Pol␦ is active but unstable. The same results were obtained when overexpression of POL3 together with POL32 was carried out. Overproduction of polymerase activity in the cleared lysates was 9-fold, but the activity also decayed upon further fractionation (Table I). Again, by Western analysis, Pol3p was found in the insoluble pellet whereas most of the overproduced Pol32p remained soluble. A gel filtration analysis of the enzyme fraction after phosphocellulose chromatography failed to identify an activity which contained both Pol3p and Pol32p but lacked Pol31p (data not shown). These data suggest that Pol3p does not form a stable active complex with Pol32p.
In contrast, overexpression of POL3 together with POL31, or of POL3 together with POL31 and POL32 resulted in a much larger increase in polymerase activity in crude extracts (Table  I). Yet, the polypeptide levels in these extracts were not higher than when Pol3p was overproduced alone or together with Pol32p ( Fig. 1). In addition, the polymerase activity from these strains remained stable during ammonium sulfate fractionation, phosphocellulose chromatography, and MonoQ HPLC ( Table I). Overexpression of POL31 together with POL32 gave an increase in the levels of these subunits, but no significant increase in DNA polymerase activity, indicating that the level of Pol3p limits the level of Pol␦ in the cell (Fig. 1, Table I).
The partially purified preparations were fractionated on a strong anion exchanger (MonoQ), which separates Pol␦ from Pol␣ and Pol⑀. In comparison to the control strain a 2.2-fold increase in Pol␦ activity was measured in the MonoQ fractions when Pol3p alone was overproduced, and a 2.6-fold increase when Pol3p and Pol32p were overproduced together ( Fig. 2A and Table I). Interestingly, two poorly separated peaks of DNA polymerase activity resulted from overproduction of Pol3p to-gether with Pol31p. Whereas the elution position of the minor peak coincides with that of Pol␦, the earlier eluting major peak represents a novel activity (Fig. 2B). The elution position is similar to that observed during fractionation of extracts from a mutant strain lacking Pol32p (7). To show that this novel peak represents a two-subunit form of Pol␦, we overproduced Pol3p together with Pol31p in a protease-deficient pol32⌬ strain. Only the early eluting peak, and no Pol␦ peak, was observed, and this activity was increased more than 10-fold in fractions from the overproducing strain (Fig. 3). For the purpose of this study we identify the two-subunit form of Pol␦ as Pol␦*.
Both Pol␦ and Pol␦* were further purified by chromatography on a strong cation exchanger. Interestingly, Pol␦* again eluted prior to Pol␦, indicating that the highly charged nature of Pol32p promotes retention of Pol␦ on either ion-exchange matrix. After the MonoS step, Pol␦ was about 95% pure and Pol␦* 50% pure (data not shown). They were further purified by gel permeation chromatography as described below.
Subunit Structure of Pol␦ and Pol␦*-Enzyme from the Mo-noS step was injected on a Superose 6 column. Pol␦ eluted at a position consistent with that of a 520-kDa complex (Fig. 4). This is in agreement with our initial studies of Pol␦ in which we noted that the apparent size of Pol␦ was Ͼ300 kDa (11). In contrast, the elution position of Pol␦* indicated a size of 180 kDa for that complex (Fig. 4). The data shown in Fig. 4 were obtained when concentrated enzyme at 0.4 mg/ml was injected onto the column. Dilution of the injected enzyme to 0.05 mg/ml did not change the respective elution positions of Pol␦ and Pol␦* (data not shown).
To obtain an estimate of the subunit stoichiometry, the peak activity fractions of the Superose 6 column were analyzed on a 10% SDS-polyacrylamide gel. The proteins were stained with Coomassie Brilliant Blue and the stained gels were digitized and quantitated. Scanning results indicate a stoichiometry for   (Fig. 5). Together with the gel filtration results, these data give as the most likely structure for Pol␦* that of a heterodimer (Pol3p⅐Pol31p), and for Pol␦ that of a hexamer: a dimer of a heterotrimer (Pol3p⅐Pol31p⅐Pol32p) 2 .
Enzymatic Activities of Pol␦ and Pol␦*-The specific activity of Pol␦ and Pol␦* was measured on activated DNA (data not shown). Calculated on a weight basis Pol␦ had a 1.3-fold higher specific activity than Pol␦*. This equates to a 3.2-fold higher specific activity for Pol␦ if calculated on a molar basis, or 1.6-fold higher if both catalytic cores in the hexameric Pol␦ are active.
The processivity of both enzymes was determined on poly(dA) 500 ⅐oligo(dT) 22 . Whereas Pol␦* incorporated 2-3 dTMP residues per binding event, Pol␦ had a processivity of 6 -12, in agreement with previous studies (Fig. 6 and Ref. 12). Upon addition of PCNA to the assay, both enzymes were fully processive. However, a much higher level of PCNA was necessary to stimulate processive DNA synthesis by Pol␦* than by Pol␦ (Fig.  6).
Pol␦* Poorly Replicates Natural DNA Templates-DNA syn-thesis by Pol␦* or Pol␦ on extended SS DNA templates is virtually completely inhibited by the presence of 75 mM NaCl in the assay. On the other hand, these conditions are optimal for replication by a complex of Pol␦ with PCNA (13). PCNA was loaded onto singly primed single stranded-binding proteincoated SS mp18 DNA by RFC and ATP. Addition of Pol␦ resulted in extremely fast and efficient DNA synthesis (Fig. 7A).  1 and 5), 0.06 pmol (lanes 2 and 6), 0.6 pmol (lanes 3 and 7), or 6 pmol of PCNA (lanes 4 and 8). The length of the poly(dA) is about 500 nt. See "Materials and Methods" for further details.
Only marginal differences in replication efficiency were observed between Pol␦ purified from a wild-type strain of yeast through a six-column procedure and Pol␦ purified from the overproducing strain through a three-column procedure. The fastest complexes complete replication of the 7,250-nt mp18 circle within 1.5 min at 37°C, a rate of more than 80 nt/s. In comparison, Pol␦* is a very inefficient enzyme, replicating only ϳ2 kb of DNA during the 20-min assay (Fig. 7A). Replication is still dependent on PCNA as no synthesis was observed in its absence. Pause sites with Pol␦* are much more pronounced than with Pol␦, indicating that sites of secondary structure form major replication barriers for this two-subunit enzyme (Fig. 7).
The poor replication efficiency of Pol␦* may be due to the frequent disassembly of replication complexes, perhaps at sites of secondary structure. In that case, rapid reassembly should be stimulated by providing excess PCNA or excess Pol␦* resulting in more efficient synthesis. Indeed, high levels of PCNA stimulated Pol␦*-mediated replication (Fig. 7B). In comparison, because Pol␦ holoenzyme is a very stable complex, a slight molar excess of PCNA over primer termini already allowed maximally efficient synthesis by Pol␦ (Fig. 7B). Similarly, a 10-fold molar excess of Pol␦* over primer termini also greatly stimulated PCNA-dependent replication of SS mp18 DNA by this enzyme (data not shown).
Reconstitution of Pol␦ from Pol␦* with Pol32p-Pol32p overproduced in E. coli is found in inclusion bodies. The protein was efficiently renatured from a 6 M urea extract. Its properties indicate that Pol32p is a homodimer (7). Pol␦* was incubated with renatured E. coli expressed Pol32p, and assayed in a holoenzyme assay on single-stranded mp18 DNA, in order to determine whether in vitro reconstitution of the three-subunit Pol␦ could be performed. The result in Fig. 8A shows that reconstitution of Pol␦ as measured by the formation of a processive holoenzyme proceeded quite efficiently. Controls, a dialyzed urea extract from E. coli cells or from cells overproducing Pol31p, show that reconstitution is specific for Pol32p.
To determine whether Pol32p would restore the dimeric structure of Pol␦ upon reconstitution, the E. coli produced subunit was purified to ϳ60% homogeneity by MonoQ HPLC, preincubated with an equimolar quantity of Pol␦*, and subjected to Superose 6 gel filtration. DNA polymerase activity, as measured on activated DNA, eluted as two poorly separated peaks at ϳ500 and ϳ200 kDa (Fig. 8, B and C). When the fractions were assayed for activity in a Pol␦ holoenzyme assay on SS mp18 DNA, the low molecular weight peak showed up as a mere shoulder to the high molecular weight peak (Fig. 8B).
An SDS-PAGE analysis showed that the peak at 500 kDa contained all three subunits of Pol␦ including Pol32p (Fig. 8C). The peak at 200 kDa also contained low amounts of Pol32p, but considering its low replication activity on SS mp18 DNA, functional Pol␦ was likely not reconsituted.

DISCUSSION
Overproduction Studies in Yeast-Overproduction of Pol␦ in yeast was easily accomplished by cloning the genes for its three subunits under control of the galactose-inducible GAL1-10 promoter (Table I). The enzyme isolated and purified from such an overproduction strain did not show marked differences with the enzyme isolated from a wild-type strain (Fig. 7). The slightly lower activity of Pol␦ isolated from a non-overproducing strain could be caused by partial inactivation of the enzyme during the laborious multistep purification procedure. (7). The availability of overproducing plasmid also allowed us to investigate the occurrence of partial Pol␦ complexes and their activities. A yeast two-hybrid analysis with Pol␦ subunits indicated a strong interaction between Pol3p and Pol31p and between Pol31p and Pol32p, and also a weak but significant interaction between Pol3p and Pol32p (7). Yet, we found no biochemical evidence for a stable complex between Pol3p and Pol32p. When both subunits were overproduced together and the extracts fractionated on a MonoQ column, the same results were obtained as when only Pol3p was overproduced: only one peak of Pol3p derived activity was identified, and this peak corresponded to the normal elution position of Pol␦ ( Fig. 2A). Although these biochemical data do not exclude the possibility of a Pol3p⅐Pol32p complex as suggested by the two-hybrid analysis, they do indicate that such a complex would be as unstable as the catalytic subunit alone. An alternative explanation of the two-hybrid results is that the Pol3p⅐Pol32p signal was the result of an indirect interaction with Pol31p serving as a bridge, i.e. Pol3p⅐Pol31p⅐Pol32p. Pol3p forms a strong complex with Pol31p to give Pol␦*, and Pol31p also forms a strong complex with Pol32p, as do the analogous subunits in S. pombe (6,7,14).
Overproduction of the catalytic subunit alone or together with Pol32p resulted in a 2-3-fold overproduction of Pol␦ from the MonoQ column ( Fig. 2A). The most simple explanation of these results is that in wild-type yeast cells, Pol31p and Pol32p are present at higher levels than Pol3p, and that, therefore, overexpression of Pol3p alone or together with Pol32p is sufficient for the observed moderate overproduction of Pol␦. In agreement with this conclusion is the observation that over- FIG. 7. Replication of SS mp18 DNA by Pol␦* and Pol␦. A, the 60-l holoenzyme replication reactions contained 0.08 pmol of primed template, 0.2 pmol of RFC, 1 pmol of PCNA, and either 0.2 pmol of Pol␦* or 0.1 pmol (expressed as hexamers) of Pol␦ (see "Materials and Methods" for details). Aliquots of 15 l were analyzed on a 1.5% alkaline-agarose gel. The control lane (lane 1) is from a reaction with Pol␦* lacking RFC. The two different forms of Pol␦ as indicated are the ones purified from wild-type strain BJ405 (7), and from the overproducing strain (Fig.  4). Size markers (in kb) are indicated at the left. B, standard replication assays contained 0.04 pmol of primed template, either 0.1 pmol of Pol␦* or 0.05 pmol (expressed as hexamers) of Pol␦ and PCNA as indicated. Assays were for 5 min at 37°C and were analyzed on a 1.5% alkaline-agarose gel. Size markers are in kb.
production of Pol31p and Pol32p did not result in increased polymerase activity (Table I).
Properties of Pol␦* and Pol␦-Most interestingly, a novel complex, Pol␦*, was isolated from the simultaneous overproduction of Pol3p and Pol31p. The differences between Pol␦* and Pol␦ are both structural and enzymatic. (i) Pol␦* is a heterodimer, whereas Pol␦ is a dimer of a heterotrimer (Figs. 4 and 5). With the knowledge that Pol32p by itself forms a homodimer, it follows that dimerization of the catalytic core must be the result of Pol32p dimerization (7). In fact, addition of Pol32p to Pol␦* restored the dimeric form of Pol␦ (Fig. 8, B  and C). (ii) The processivity of Pol␦ alone on poly(dA)⅐oligo(dT) is higher than that of Pol␦* (Fig. 6). (iii) In the presence of PCNA both DNA polymerases are fully processive on poly(dA)⅐ oligo(dT). However, much more PCNA is required to make processive complexes with Pol␦* than with Pol␦ (Fig. 6). PCNA rapidly dissociates from linear DNA and is only stabilized onto the DNA by interaction with the polymerase (15)(16)(17). Therefore, PCNA trimers loaded by diffusion onto poly(dA)⅐oligo(dT) rapidly slide off the DNA as they fail to be anchored much less efficiently by Pol␦* than by Pol␦. However, those few PCNA clamps which do form a complex with Pol␦*, replicate processively. (iv) Pol␦* holoenzyme (i.e. the complex of polymerase, PCNA, and RFC) is much less efficient than Pol␦ holoenzyme in the replication of natural DNA templates (Fig. 7). Whereas Pol␦ holoenzyme replicated SS mp18 DNA with high processivity, replication by Pol␦* holoenzyme showed frequent pausing. Those replication defects were partially suppressed by a large molar excess of PCNA or Pol␦*, suggesting that pausing led to frequent holoenzyme disassembly, and subsequent reassembly was stimulated by excess PCNA and Pol␦* (Fig. 7B, data not shown). As the PCNA⅐Pol␦* complex is processive on poly(dA)⅐oligo(dT) which lacks secondary structure, it is likely that the replication defects of Pol␦* holoenzyme on SS mp18 DNA are due to the extensive secondary structure of this template. In conclusion, although the interactions between PCNA and the Pol3p and/or Pol31p subunits are essential for establishing a productive PCNA-polymerase complex, additional interactions between PCNA and Pol32p may stabilize this complex, particularly during replication of secondary structures in the DNA template. Alternatively, or in addition, the presence of the third subunit itself may stabilize the holoenzyme complex.
Yeast Pol␦* Is Comparable to Human Pol␦-The two-subunit yeast Pol␦* is structurally and functionally analogous to mammalian Pol␦ (1,18). Like Pol␦*, mammalian Pol␦ is purified as a heterodimer. Depending on the PCNA levels, the synthetic rate of Pol␦* holoenzyme varies from 1.5 to 15 nt/s, much less than the rate of the three-subunit Pol␦ at ϳ100 nt/s (Fig. 7) (19). The latter rates are comparable to in vivo rates of fork movement in yeast (20,21). The range of synthetic rates of human or bovine Pol␦ holoenzyme in analogous replication reactions is 2-10 nt/s (e.g. see Refs. 22 and 23). As with yeast Pol␦* holoenzyme, replication by mammalian Pol␦ holoenzyme is also prone to pausing. It appears that the large difference in replication efficiency between yeast and mammalian Pol␦ is largely or completely accounted for by the presence of the third subunit in the yeast enzyme. Is this subunit also present in mammals? As Pol32p and Cdc27, the functional S. pombe homologue of Pol32p, show only minimal sequence similarity, it may not be possible to clone or identify this putative mammalian subunit based on sequence comparison considerations. The prediction would be that human Pol␦ containing the third subunit would be more processive. SV40 might be an attractive assay system for this third subunit as the assay can be carried out in crude extracts (24). Unfortunately, as the synthetic rate of the SV40 fork is limited at 3 nt/s by the rate of the T antigen helicase, it may not be possible to functionally detect the presence of the third subunit, based on rate considerations only (22).
Is Pol␦ a Dimer at the Fork?-The observation that Pol␦ has a dimeric catalytic core immediately suggests the notion that this enzyme is also a functional dimer at the replication fork. Replication of both the leading and the lagging strand by Pol␦ is the favored model for SV40 (25,26). However, in yeast there are several indications that Pol⑀ plays a major role at the replication fork. First, Pol⑀ is essential for yeast cell growth and the phenotypes of temperature-sensitive Pol⑀ mutants indicate that the enzyme is required for the elongation phase of DNA replication (27,28). Second, in vivo cross-linking studies place Pol⑀ at or near the fork (29). Third, the mutator pheno-FIG. 8. Reconstitution of Pol␦ from Pol␦* and Pol32p. A, cell pellets from lysed E. coli cells or from cells overexpressing POL31 or POL32 were extracted with a buffer containing 6 M urea, and the extract was dialyzed stepwise to a no urea buffer (7). Pol␦* (0.5 g) was incubated with 5 l of dialysate, which contained either no overproduced protein (lanes 1 and 5) or approximately 1 g of Pol31p (lanes 2 and 6) or 2 g of Pol32p (lanes 3 and 7), in a total volume of 20 l of 40 mM Hepes-NaOH, pH 7.5, 10% glycerol, 1 mM EDTA, 0.02% Nonidet P-40, 0.2 M NaCl, 5 mM DTT, 5 mM NaHSO 3 , and 2 M each of leupeptin and pepstatin A. After an incubation for 1 h on ice, standard mp18 DNA replication assays were carried out in an assay containing 0.5 pmol of PCNA, 2 l (0.2 pmol) each of the reconstituted enzymes (lanes 1-3 and 5-7), or 0.05 pmol (expressed as hexamers) of Pol␦ (lanes 4 and 8). Aliquots taken after 2 min (lanes 1-4) and 10 min (lanes 5-8) were analyzed on a 1.5% alkaline-agarose gel. B, the dialyzed Pol32p obtained from a 1-liter E. coli overproduction scale was further purified by MonoQ HPLC in buffer C using a NaCl gradient from 100 to 500 mM. The subunit eluted broadly between 200 and 350 mM NaCl. The purest fraction (ϳ60% pure) eluting at 300 mM NaCl (100 l, 12 g of protein representing ϳ7 g ϭ ϳ0.18 nmol monomer of Pol32p) was incubated with 50 l (23 g ϭ 0.13 nmol of heterodimer) of Pol␦* for 1 h at 0°C and analyzed on a Superose 6 column. Fractions of 0.3 ml were collected and analyzed for DNA polymerase activity on activated DNA (E) and for holoenzyme activity on mp18 DNA (q). C, proteins in the peak fractions were separated by 10% SDS-PAGE and visualized by silver staining. Fractions 26, 30, and 34 represent elution volumes of 11.8, 12.8, and 13.8 ml, respectively. The elution positions of markers is as indicated. For details see "Materials and Methods" and the legend to Fig. 4. type of a proofreading-deficient mutant of Pol⑀ suggests a replication function for Pol⑀. In particular, the multiplicative relationship of spontaneous mutation rates between Pol⑀ exonuclease-deficient mutants and mismatch repair mutants indicates that these pathways act sequentially, i.e. that the proofreading function of Pol⑀ is required during DNA replication and prior to mismatch repair (30). Furthermore, the observed spectrum of 6-N-hydroxylaminopurine-induced mutations in strains defective for the proofreading exonuclease of either Pol␦ or Pol⑀ indicate that the respective exonuclease functions of these DNA polymerases correct the analog induced DNA replication errors on opposite DNA strands (31). By extension, if we assume that the DNA synthetic and proofreading functions of a DNA polymerase are tightly coupled, these latter data indicate that Pol␦ and Pol⑀ replicate opposite strands of the fork (32). Taken in total, the various data pointing to the respective replication functions of these two DNA polymerases remain inconclusive and perhaps even contradictory.
Function of Pol32p-The Pol32p subunit of Pol␦ has at least three functional domains, a basic structural domain for interaction with Pol31p, an organizational domain for homodimer formation thereby promoting dimerization of the catalytic core, and a domain which interacts with PCNA (7). None of these functions is essential for yeast cell growth. Recently, we described the properties of a PCNA mutant, pcna-79, with mutations in conserved residues in the interdomain connector loop (I126A, L128A) (33). The mutant PCNA fails to interact with the Pol32p subunit and the in vitro replication properties of a pcna-79 containing Pol␦ holoenzyme are quite similar to those of the Pol␦* holoenzyme described here, suggesting that the observed in vitro phenotype of Pol␦* holoenzyme may be due to the loss of a PCNA interaction site. However, mutant yeast cells containing the pol30-79 mutation differ in phenotype from mutants deleted for the POL32 gene. Whereas both mutants are sensitive to hydroxyurea indicative of replication defects, pol32⌬ mutants are cold-sensitive for growth but pol30-79 mutants are not. Furthermore, pol30-79 is a mutator and pol32⌬ an antimutator with a defect in damage-induced mutagenesis (7,33). Loss of Pol32p function shows a much more severe growth defect than loss of the PCNA-Pol32p interaction. Possibly, the loss of dimerization function of Pol32p rather than PCNA interaction leads to the observed conditional lethal phenotype in pol32⌬ mutants. If the function of Pol␦ as a dimeric enzyme is important for yeast replication, it appears that other factors contribute to the stabilization of a dimeric replisome, and, therefore, loss of Pol32p-Pol32p interactions can be toler-ated. A comprehensive mutational analysis of the POL32 gene is required to address these questions appropriately.