Human Pol ζ purified with accessory subunits is active in translesion DNA synthesis and complements Pol η in cisplatin bypass

Significance

Although human DNA polymerase ζ (Pol ζ) is essential for DNA replication and cell proliferation, difficulties purifying active Pol ζ have hindered its biochemical characterization. We report here the first purification of an active form of human Pol ζ holoenzyme composed of Rev3, Rev7, PolD2, and PolD3, which opens up the possibility for detailed biochemical and structural studies of this essential enzyme. Based on genetic data, it has been postulated that two specialized DNA polymerases are needed for successful translesion synthesis. We show here that human Pol η inserts a nucleotide opposite the lesion, followed by Pol ζ extending the DNA primer; thus, the two complement each other to fully bypass the cisplatin cross-link.

Abstract

DNA polymerase ζ (Pol ζ) is a eukaryotic B-family DNA polymerase that specializes in translesion synthesis and is essential for normal embryogenesis. At a minimum, Pol ζ consists of a catalytic subunit Rev3 and an accessory subunit Rev7. Mammalian Rev3 contains >3,000 residues and is twice as large as the yeast homolog. To date, no vertebrate Pol ζ has been purified for biochemical characterization. Here we report purification of a series of human Rev3 deletion constructs expressed in HEK293 cells and identification of a minimally catalytically active human Pol ζ variant. With a tagged form of an active Pol ζ variant, we isolated two additional accessory subunits of human Pol ζ, PolD2 and PolD3. The purified four-subunit Pol ζ4 (Rev3–Rev7–PolD2–PolD3) is much more efficient and more processive at bypassing a 1,2-intrastrand d(GpG)-cisplatin cross-link than the two-subunit Pol ζ2 (Rev3–Rev7). We show that complete bypass of cisplatin lesions requires Pol η to insert dCTP opposite the 3′ guanine and Pol ζ4 to extend the primers.

DNA polymerase ζ (Pol ζ), composed of the catalytic Rev3 and accessary Rev7 subunits, is an error-prone DNA translesion polymerase that causes both spontaneous and DNA damage-induced mutagenesis (1, 2). More than two-thirds of the 1,504 residues in yeast Rev3 share sequence homology with all B-family DNA polymerases, including Pols α, δ, and ε, which are responsible for the bulk of high-fidelity genomic replication in eukaryotes (3). Unlike the typical B-family polymerases, Pol ζ lacks an intrinsic 3′–5′ exonuclease activity and thus has no proofreading function (2). Human homologs of REV3 (REV3L) and REV7 (MAD2L2; hereafter referred to as REV7) genes were identified shortly after yeast Pol ζ was characterized. Human Rev3 contains 3,130 residues and is twice as large as the yeast counterpart (4). Human and yeast Rev7 are homologous (5) and bear sequence similarity to the mitotic checkpoint proteins Mad2 (6). Unlike Saccharomyces cerevisiae REV3 and REV7 genes, which are nonessential and whose knockout leads only to a decreased rate of damage-induced mutagenesis (7, 8), Rev3l knockout in mice is embryonic-lethal (9), and mouse Rev3l^−/− embryonic stem cells are not viable (10, 11). Human and mouse cell cultures obtained from conditional Rev3l knockout show genome instability and growth defects without an external challenge of DNA damage (12–14). DNA pol ζ is apparently essential for normal cell proliferation and embryogenesis in mammals.

Translesion synthesis (TLS) and DNA-damage-induced mutagenesis are the best-characterized functions of Pol ζ. Absence of the yeast REV3 gene leads to sensitivity to UV light and intrastrand and interstrand cross-linking agents (2, 15). DNA Pol ζ has been shown to induce multiple base substitutions as well as more complex mutations in yeast (7, 16, 17) and may contribute to hypermutation in Ig genes in mammals (18, 19). The TLS function of DNA Pol ζ has been implicated in its role of mediating resistance to platinum-based chemotherapies (20–22). Owing to the conservation of B-family DNA polymerases, a distorted DNA template base is unlikely to be accommodated in the active site of DNA Pol ζ. In fact, yeast DNA Pol ζ is unable to insert a nucleotide opposite either a cis–syn thymine or a 6-4 photoproduct (23). Genetic data indicate that a complete lesion bypass event may require two TLS DNA polymerases (24)—one for nucleotide incorporation opposite a lesion (insertion step) and the other for the subsequent primer extension (extension step). The insertion step of TLS is often accomplished by a Y-family polymerase, whose active site is uncommonly large, solvent-exposed, and flexible (25). Studies of another B-family TLS DNA polymerase from Escherichia coli (Pol II) show that it efficiently extends a DNA primer after a lesion by looping out the damaged DNA template strand, leading to frameshift and mixed-type mutations (26).

In budding yeast, REV3 has been shown to be epistatic with POL32, a subunit of DNA Pol δ. Inactivating either REV3 or POL32 leads to reduced spontaneous mutagenesis (27–29). As with all eukaryotic B-family DNA polymerases, Rev3 contains a Cys-rich C-terminal domain (CTD) (30–33), which forms a zinc-finger domain followed by a [4Fe–4S] cluster (34). In Pol α, δ, and ε, each CTD interacts with its specific accessary subunits (32, 35). Recently, three groups have independently shown that the [4Fe–4S] cluster of yeast Rev3 interacts with Pol31 and Pol32 subunit (36), thus forming an stoichiometric four-subunit Pol ζ (Pol ζ4; Rev3–Rev7–Pol31–Pol32) (23, 37, 38). Baranovskiy et al. further showed that the CTDs of human Pol ζ and δ share the same accessary subunits p50 and p66, homologs of yeast Pol31 and Pol32, respectively (37). The interaction between yeast Rev3 and Pol31 is reported to be direct, and Pol32 is essential to stabilize Pol31 and, furthermore, via its interactions with proliferating cell nuclear antigen (PCNA), recruits and activates Pol ζ to carry out TLS (38). The catalytic activity of yeast Pol ζ is improved by the presence of Pol31 and Pol32 (23, 38).

Purification and characterization of Pol ζ has so far been limited to the yeast protein. Perhaps because of its large size, mammalian Pol ζ has not been purified for biochemical characterization. To overcome this roadblock, we coexpressed human REV3L and REV7 in mammalian cells in culture. Initially, very low expression level and heterogeneity was encountered, but these problems were solved by targeted deletion of various internal segments of human REV3L. We succeeded in purifying an active two-subunit form of human Pol ζ (Pol ζ2). By differential pull-down experiments using Pol ζ2 variants with and without the CTD of Rev3, we isolated two CTD-dependent Pol ζ accessary subunits, PolD2 and PolD3. We report here purification of an active form of human four-subunit Pol ζ4 and the collaboration of two TLS polymerases, Pol η and Pol ζ, in lesion bypass.

Results

Coexpression and Purification of Human Rev3 and Rev7.

Two major insertions in human Rev3 separate three highly conserved domains: the N-terminal domain (NTD; 1–333 aa), the Rev7-binding domain (R7B; 1,888–1,943 aa), and the C-terminal polymerase domain (Pol; 2,276–3,130 aa) (ref. 4; Fig. 1A). Most of the inserted sequences (∼1,500 aa) are predicted to be random coil interspersed with low complexity segments (bioinf.cs.ucl.ac.uk/psipred). Scrutinizing the primary structure of the inserted sequences, we identified a positively charged domain (PCD; 960–1,200 aa), which is ubiquitously present between the NTD and R7B among vertebrate Rev3 orthologs. Accordingly, we constructed expression vectors of the full-length human REV3L and six variants with deletions of the predicted unstructured sequences that surround the three conserved domains and the PCD (Fig. 1A). His₈ and maltose binding protein (MBP) tags were placed in tandem on the N terminus of Rev3 or Rev7 to increase protein solubility and simplify protein purification. The seven REV3L constructs were individually cotransfected with REV7 into HEK293 cells for transient protein expression. Pol ζ2 was partially purified from harvested cells by using an amylose affinity column (Fig. 1B and Fig. S1). When full-length Rev3 was MBP tagged, only trace amounts of a degraded MBP-fusion protein of ∼110 kDa was detected by Western blot (Fig. S1A). Even with a Rev3 deletion variant, the yield of Pol ζ2 was 10 times greater when Rev7 was tagged instead of Rev3 (Fig. S1B). Therefore, we chose to tag Rev7 and not Rev3.

Fig. 1.

Determination of a minimal Rev3 that forms active human Pol ζ. (A) Alignment of the conserved domains between yeast and human Rev3. The NTD is shaded in blue, R7B in yellow, and the Pol in red. Domain boundaries are indicated by residue numbers. The positively charged domain (PCD) shaded in green is unique to vertebrate Rev3. In human Rev3 deletion constructs TR1, TR3, TR4-2, TR4-3, TR4-5, and TR5, predicted unstructured portions with boundaries marked by residue numbers were deleted and replaced by two amino acids as indicated. (B) SDS gel stained by Coomassie Blue R-250 of purified Rev3–Rev7 (Pol ζ2) variants. Each variant partially purified by amylose column was loaded at an equal amount (2.5 μg) and labeled according to the Rev3 construct. MBP–Rev7 was more abundant than Rev3 in each case. (C) DNA synthesis by each of six Pol ζ2 variants. Equal proportions of Pol ζ2 proteins as loaded in the SDS gel (in B) were used in the DNA synthesis assay. The DNA primers were 5′-FAM Fluorescein-labeled. (D) Activity comparison of Pol ζ2 variants. The amount of Rev3 in each lane in B (representing active Pol ζ2) and the amount of the full-length DNA products (43 and 44 nt) in C were quantified based on the band intensity. Data for the full-length Rev3 are shown in Fig. S1. The specific activity of each Pol ζ2 variant (the amount of DNA product divided by the amount of Rev3 protein) was calculated and then normalized to that of TR4-2.

Coexpression and purification of the full-length Rev3 and MBP–Rev7 yielded low amounts and low purity of human Pol ζ (Fig. S2A). In addition to the 353-kDa band of the full-length hRev3, an equal intensity band of ∼300 kDa was observed, which likely corresponds to a large fragment of Rev3 (see below). With a single internal deletion of residues 527–1,800 in Rev3 (TR1), the yield of Pol ζ increased by ∼10-fold (Fig. 1 A and B). However, the specific polymerase activity of TR1 on a normal DNA template–primer pair was ∼2% that of full-length Pol ζ (Fig. 1 and Fig. S2). Inclusion of the positively charged patch (1,042–1,254 aa) and alteration of boundaries surrounding the R7B domain (TR3, TR4-2, and TR4-3) resulted in an approximately eightfold specific-activity increase of TR4-2 relative to TR1 without reducing the protein yield (Fig. 1). Comparison among TR1, TR3, TR4-2, and TR4-3 constructs suggested that inclusion of amino acids 1,588–1,800 and deletion of amino acids 2,048–2,217 enhanced the protein yield and specific activity. The most significant increase in the Pol ζ-specific activity was achieved by the addition of residues 900–1,041 (TR4-5 and TR5) (Fig. 1), which restored the specific catalytic activity of TR5 (a total of 2,358 aa) to that of the full-length Rev3 (3,130 aa). However, this activity increase was accompanied by a significant decrease in protein yield and the appearance of a proteolyzed Rev3 band (Fig. 1B). The size difference between the expected Rev3—whether full-length, TR4-5, or TR5—and the proteolyzed bands on SDS gels appeared to be constant. We therefore estimate that the cleavage site is around residue 500 and suspect that the presence of residues 900–1,041 somehow alters the Rev3 structure to expose this protease cleavage site. For the remaining Pol ζ studies reported here, we focus on the TR4-2 variant (2,113 of 3,130 aa total), which produces a single Rev3 band on SDS gels and an active Pol ζ2 that has the same nucleotide preference and generates the same DNA products as the full-length Rev3 (Fig. S2).

Identification of Human DNA Pol ζ Accessory Subunits.

The conserved Cys-rich CTDs of B-family DNA polymerases α, δ, and ε each interact with unique accessory subunits. To determine whether the CTD of human Rev3 (3,016–3,130 aa) also interacts with its own accessory proteins, we constructed a CTD-deletion clone of TR4-2 (TR4-2ΔCTD) and cotransfected expression vectors of TR4-2 with or without CTD and His₈–MBP–Rev7 into HEK293 cells (Fig. 2A). To ensure that the amount of Rev3 and Rev7 did not greatly exceed endogenous Pol ζ cofactors, we reduced the DNA amounts used for transfection. Without assuming which proteins might interact with the CTD of Pol ζ, we fractionated the total HEK293 cell lysate with an amylose pull-down followed by a cation exchange column. The eluents were treated with PreScission protease to separate Rev7 from the His₈–MBP tag and analyzed on a SDS gel with Coomassie blue stain (Fig. 2B). The presence of the Rev3 CTD resulted in two additional pull-down protein bands. Band 1 (∼65 kDa) was completely absent in the lane without the Rev3 CTD, and band 2 (∼50 kDa) ran very closely to a protein band (band x) present in the lane without the Rev3 CTD (Fig. 2B). With >55% of each protein band sequenced by mass spectrometry, bands 1 and 2 were identified as human PolD3 (p66) and PolD2 (p50), homologs of yeast Pol32 and Pol31, respectively. Band x was identified to be α and β tubulins. The other protein band (band y) pulled down with or without the Rev3 CTD was ∼75 kDa and is too small to be Rev1, which is known to associate weakly with Rev7 (39–42). Based on our prior experience with protein expression in HEK293 cells, band y is likely a protein chaperone. We conclude that, like yeast Pol ζ, human Pol ζ is a four-subunit DNA polymerase and shares two accessory subunits with DNA Pol δ.

Fig. 2.

Identification of human Pol ζ accessary subunits. (A) A diagram of Pol ζ and potential accessary subunits. (B) Pull-down of Rev3 accessary subunits from HEK293 cells. The TR4-2 variants of Rev3 with or without CTD were constructed and coexpressed with MBP–Rev7 as baits to pull down Pol ζ2-interacting proteins. Bands 1 and 2 were purified only in the presence of the CTD. Band 2 overlapped with band x, which was copurified with Rev3 with or without the CTD. (C) Mass spectrometry analysis identified bands 1 and 2 as PolD3 (p66) and PolD2 (p50), respectively, and band x as tubulins.

PolD2 and PolD3 Enhance the Catalytic Efficiency and Processivity of Pol ζ.

To investigate how PolD2 and PolD3 influence the stability and catalytic activity of human Pol ζ, we prepared Pol ζ2 and Pol ζ4 by cotransfection of two or four relevant expression vectors of TR4-2, His₈–MBP–Rev7, PolD2 (p50 with or without an MBP tag), and PolD3 (p66 with or without a GST tag) (Fig. 3 A–C) into HEK293 cells. To ensure that no endogenous PolD2 and PolD3 copurified with Pol ζ2, a TR4-2ΔCTD clone was used to generate Pol ζ2. For the negative control in enzymatic assays, we generated and purified inactive Pol ζ4⁻ protein by mutating two catalytically essential aspartates in Rev3 to asparagines (D2614N and D2783N). Differently tagged Pol ζ4 proteins (MBP-tagged p50 or GST-tagged p66) were purified by amylose or amylose and glutathione affinity columns, followed by cation-exchange chromatography. Pol ζ2, ζ4, and ζ4⁻ were equally well expressed and equally stable (Fig. 3). Interestingly, PolD2 (p50) or PolD3 (p66) that was untagged—and thus might suffer loss during the affinity purification—was present in less than stoichiometric amounts in the purified four-subunit complex Pol ζ4 (Fig. 3 B and C). The varying ratios of PolD2 and PolD3 in the human Pol ζ4 preparations when different subunits were tagged suggest that PolD3 may interact with human Rev3 or Rev7 independent of the PolD2–CTD interaction (23, 37, 38).

Fig. 3.

DNA synthesis by human Pol ζ2 and ζ4. (A) Diagrams of Pol ζ4 (Left), ζ2 (Center), and ζ4⁻ (Right), which were inactivated by mutations in the polymerase active site (D2614N/D2783N). (B) Protein preparation (Left) and DNA synthesis (Right) assay of the Pol ζ4 with both Rev7 and PolD2 MBP-tagged. The amount of MBP–PolD2 (p50) exceeded the untagged PolD3 (p66), which ran slightly faster than MBP–Rev7. With equal amount of Rev3, Pol ζ4 was much more efficient than Pol ζ2 in producing the full-length DNA products. (C) Protein preparation (Left) and DNA synthesis (Right) assay of the Pol ζ4 with GST-tagged p66. The amount of GST–PolD3 (p66) exceeded the untagged PolD2 (p50). The DNA synthesis activities of the two differently tagged Pol ζ4 were comparable, indicating that all four subunits were essential, but the reduced amount of an accessary subunit might limit the apparent activity of Pol ζ4. (D) Measurement of K_M and k_cat of Pol ζ2 and ζ4 in dATP incorporation opposite a normal dT template base (P25–T44 DNA). (E) Comparison of the Pol ζ2 and ζ4 processivity. Dilution series of Pol ζ2 (35, 70, and 140 nM) and Pol ζ4 (13, 35 and 70 nM) were used in the primer extension assay. Both the primer use and the full-length extension (42∼44 nt) products were quantified and are indicated below each lane.

To compare the Pol ζ catalytic activities, equimolar concentrations of Pol ζ2, ζ4, and ζ4⁻ proteins were tested in normal DNA synthesis assays (Fig. 3 B–D). Although negligible, residual DNA synthesis activity was detected with the Pol ζ4⁻ protein (Fig. 3 B–E), which could be due to a trace contaminant with active DNA polymerases—for example, Pol δ or Rev1 copurifying via interaction with PolD2 or Rev7. Even with PolD2 or PolD3 in substoichiometric amounts relative to Rev3 (TR4-2), Pol ζ4 exhibited higher k_cat and lower K_M than Pol ζ2 (Fig. 3D) and was nearly 30 times more efficient than Pol ζ2 in catalyzing DNA synthesis in the absence of a lesion.

The presence of PolD2 and PolD3 accessory subunits also increased the processivity of human Pol ζ4 on DNA substrates. Full-length DNA primer-extension products (42- to 44-nt bands) were prominent with Pol ζ4 but very weak with Pol ζ2 (Fig. 3 B and C). To semiquantitatively assess the processivity of Pol ζ4 and Pol ζ2, the primer extension assay was performed with a range of concentrations of the two protein complexes, and primer use and product distributions were then compared (Fig. 3E). At the lowest concentration of Pol ζ4 and highest concentration of Pol ζ2, for which nearly 50% of DNA primers were extended in each case, Pol ζ4 produced at least five times more full-length DNA products than Pol ζ2. The increased processivity was likely due to enhanced DNA binding by human Pol ζ4, which may also explain the reduced K_M.

Pol ζ and Pol η Synergistically Bypass an Intrastrand Cisplatin Cross-Link.

Human Pol ζ has been implicated in resistance to platinum-based chemotherapies (20–22), and knockdown of REV3L sensitizes malignant cells to cisplatin treatment (21, 43). We compared the catalytic activity and accuracy of Pol ζ4 in synthesizing DNA in the presence or absence of a 1,2-intrastrand d(GpG)-cisplatin cross-link (abbreviated as Pt-GG hereafter). The relative catalytic efficiencies of Pol ζ4 in nucleotide incorporation opposite a cross-linked Pt-GG and in primer extension after the lesion were separately assayed at each step by using four primers of different lengths paired with two template strands of the same sequence—one with Pt-GG and the other without (Fig. 4A). Pol ζ4 was inactive when encountering the 3′ G of Pt-GG (the first of two cross-linked GG) and failed to insert any nucleotide in the presence of individual or all four nucleotides. Pol ζ4, however, accurately inserted a C opposite the 5′ G of Pt-GG (the second of the cross-linked GG). The efficiency of primer extension after Pt-GG by Pol ζ4 increased step by step with increased k_cat and decreased K_M (Table 1). With a normal base pair between the template and primer after the Pt-GG lesion, primer extension by Pol ζ4 reached 10–15% of the catalytic efficiency and >95% of k_cat observed in the absence of Pt-GG (Table 1). A contaminant 3′-exonuclease activity was observed when the P26 primer was paired with the Pt-GG template (Fig. 4B), and the resulting P26 degradation was lesion dependent and Pol ζ4 activity independent (Fig. S3). Consistent with its error-prone reputation, Pol ζ4 misincorporated A, T, or G opposite a template G in the normal or Pt-GG DNA (Fig. 4B), although the degree of nucleotide discrimination will require further quantitative studies.

Fig. 4.

Pol ζ4 and Pol η carry out TLS synergistically bypassing a Pt-GG lesion. (A) Diagrams of DNA primer and template used in the assay. The four primers allow separate analyses of nucleotide insertions opposite the Pt-GG dimer and primer extension 1 or 2 nt beyond the lesion. The template strands contain either cisplatin cross-linked Pt-GG (in red) or undamaged DNA of the same sequence. (B) Primer extension of undamaged and Pt-GG DNA template by Pol ζ4 in the presence of each (A, T, G, or C) or all four nucleotides (labeled as 4). Percentage of each primer extended in the presence of a correct dNTP or mixed four nucleotides was quantified and is indicated. (Upper) Insertion step. (Lower) Extension step. (C) Pol ζ4 and Pol η were synergistic in DNA synthesis. (Upper) With normal DNA template, Pol η was efficient in nucleotide incorporation but had poor processivity. (Lower) In the presence of a Pt-GG lesion, Pol ζ4 failed to insert opposite the 3′ G of the Pt-GG, whereas Pol η failed to extend after the Pt-GG lesion. Together, Pol ζ4 and Pol η resulted in 10% of full-length product.

Table 1.

Catalytic efficiency of Pol ζ4 in normal DNA synthesis and Pt-GG bypass

Primer	Template	Incoming nucleotide	kcat, min−1	KM, μM	kcat/KM, M−1⋅min−1	Relative efficiency, % of kcat/Km	Relative rate, % of kcat
P24 (second insertion)	Undamaged	dCTP	0.22 ± 0.01	1.50 ± 0.12	144.7 x 103	100	100
	Pt-GG		0.09 ± 0.01	55.8 ± 10.4	1.7 x 103	1.2	43.3
P25 (first extension)	Undamaged	dATP	0.31 ± 0.01	1.17 ± 0.08	261.5 x 103	100	100
	Pt-GG		0.15 ± 0.01	35.6 ± 4.2	4.1 x 103	1.6	47.7
P26 (second extension)	Undamaged	dGTP	0.22 ± 0.01	0.82 ± 0.09	268.3 x 103	100	100
	Pt-GG		0.21 ± 0.01	8.80 ± 0.97	24.1 x 103	9.0	96.4

In our previous studies of Pt-GG bypass, we found that human Y-family Pol η (1-432aa) is both accurate and efficient in inserting a C opposite the first G of a Pt-GG lesion, but its catalytic efficiency decreases steeply in subsequent primer extension (44). The opposite trends of Pol ζ4 and Pol η in catalytic efficiency of bypassing Pt-GG suggest that they may complement each other and synergize to carry out TLS. Indeed, when the two polymerases were combined, Pt-GG bypass starting opposite the lesion was far more efficient and more complete than when using either polymerase alone (Fig. 4C). Notably, Pol η had low processivity on both damaged and normal DNA, and Pol ζ4 had lower catalytic activity but higher processivity than Pol η. Pol η was able to assist Pol ζ4 to overcome the Pt-GG block (with the P23 primer) as well as stalling in the middle of primer extension opposite undamaged DNA (Fig. 4C).

Discussion

We succeeded in expressing and purifying an active form of human Pol ζ, which enabled us to carry out the previously impossible in vitro characterization of Pol ζ. Mammalian Rev3 is twice as large as the yeast homolog. Among the 1,500 residues (500∼1,800 and 2,000∼2,200 aa) of Rev3 that are unique in higher eukaryotes, we identified a 340-residue highly positively charged domain (residues 960–1,200), which contains low-complexity sequence dominated by Lys (17.8%), Ser (10%), and Arg (8.3%). Its inclusion enhances the specific activity of DNA synthesis by Pol ζ. The exact amino acid sequence of PCD is not highly conserved, but the Lys- and Arg-rich character is conserved from invertebrates to mammals, suggesting that charge–charge interactions may lead to increased DNA binding by Pol ζ. Our deletion analysis also revealed that the region between R7B (Rev7 binding) and the polymerase catalytic domain (residues 2,048–2,217) negatively impacts the polymerase activity of Pol ζ2 (Fig. 1). The remaining 770 residues, which can be removed from human Rev3 without reducing the polymerase activity of Pol ζ2 as shown by our TR-5 construct, may play a role in interactions with other factors to regulate the activity of Pol ζ and its recruitment to a replication fork.

Based on studies of yeast Pol ζ and the CTD of human Rev3, it is not surprising that human PolD2 and PolD3, the equivalents of yeast Pol31 and Pol32, are accessory subunits of human Pol ζ. Our pull-down experiments using whole-cell lysates of HEK293 cells, however, were conducted without the assumption of PolD2 and PolD3 involvement. Only PolD2 and PolD3 copurified with Rev3 (TR4-2) and Rev7 after stringent washing with a 1.0 M NaCl solution, and Rev1, a known cofactor of Pol ζ via interactions with Rev7 (41, 42, 45), was not detected (Fig. 2B). PolD2 and PolD3 must have stronger interactions with Rev3 and Rev7 than Rev1. In yeast, Pol31 directly interacts with the CTD of Pol ζ and Pol δ, and Pol32 interacts only with Pol31 (38). However, the recent EM study of yeast Pol ζ4 placed Pol32 proximal to Rev7 (46). Potential interactions between human PolD3 (yeast Pol32 homolog) and Rev7 may explain our observation that PolD3 remains a component of Pol ζ4, even when PolD3 is in molar excess of PolD2 (Fig. 3C). Unless PolD2 and PolD3 are limiting in vivo, both Pol δ and ζ may exist as a four-subunit complex without competing for the accessory units.

In human Pol ζ4, the presence of PolD2 and PolD3, even at submolar ratios to Rev3 (TR4-2), dramatically increased the efficiency (k_cat/K_M) and processivity of DNA synthesis (Fig. 3D). The effect of decreased K_M and increased processivity suggests that PolD2 and PolD3 significantly enhance association of Pol ζ with DNA substrates. The cocrystal structure of human PolD2 and the NTD of PolD3 has been determined (35) and shows a primarily negatively charged molecular surface without obvious features for DNA binding. It is possible that the unstructured, but positively charged, region in PolD3 that is not included in the crystal structure may play a crucial role in enhancing the Pol ζ catalytic activity and processivity. In contrast, inclusion of two accessory subunits in yeast Pol ζ4 primarily increases the k_cat by twofold to threefold without an appreciable change of K_M or processivity (23). This difference between yeast and human Pol ζ4 may result from low similarity between human PolD3 and yeast Pol32 accessory subunits.

Pol ζ and Pol η are synergistic in improving the efficiency and processivity of DNA synthesis both in the absence and presence of a DNA lesion (Fig. 4C). In the absence of direct interactions between Pol ζ and Pol η, the synergy most likely takes place by passing DNA substrate between the two polymerases. Pol η and Pol ζ appear to complement each other in catalytic efficiency and processivity in both TLS, as demonstrated here and in maintenance of common fragile sites in human genome as reported (14, 47). However, without necessary regulation of their access to DNA replication forks, the substantial efficiency and processivity of these two low-fidelity DNA polymerases could lead to a high frequency of mutagenesis. We expect that stringent down-regulation of Pol ζ may keep this essential polymerase engaged only when necessary for lesion bypass and cell proliferation.

Materials and Methods

Human Pol ζ2 and ζ4 with His₈–MBP-labeled Rev7 were expressed in transiently transfected HEK293 cells and purified by using affinity pull-down, ion-exchange, and size-exclusion chromatography. The catalytically inactive Pol ζ4⁻ mutant was generated by using QuikChange (Stratagene). Catalytic activities of Pol ζ variants were measured with Fluorescein (6-FAM)-labeled primers as described (26). A full description of materials and methods can be found in SI Materials and Methods.