The POLD3 subunit of DNA polymerase δ can promote translesion synthesis independently of DNA polymerase ζ

The replicative DNA polymerase Polδ consists of a catalytic subunit POLD1/p125 and three regulatory subunits POLD2/p50, POLD3/p66 and POLD4/p12. The ortholog of POLD3 in Saccharomyces cerevisiae, Pol32, is required for a significant proportion of spontaneous and UV-induced mutagenesis through its additional role in translesion synthesis (TLS) as a subunit of DNA polymerase ζ. Remarkably, chicken DT40 B lymphocytes deficient in POLD3 are viable and able to replicate undamaged genomic DNA with normal kinetics. Like its counterpart in yeast, POLD3 is required for fully effective TLS, its loss resulting in hypersensitivity to a variety of DNA damaging agents, a diminished ability to maintain replication fork progression after UV irradiation and a significant decrease in abasic site-induced mutagenesis in the immunoglobulin loci. However, these defects appear to be largely independent of Polζ, suggesting that POLD3 makes a significant contribution to TLS independently of Polζ in DT40 cells. Indeed, combining polη, polζ and pold3 mutations results in synthetic lethality. Additionally, we show in vitro that POLD3 promotes extension beyond an abasic by the Polδ holoenzyme suggesting that while POLD3 is not required for normal replication, it may help Polδ to complete abasic site bypass independently of canonical TLS polymerases.


INTRODUCTION
Pol␦ and Pol⑀ are the major eukaryotic replicative polymerases. Pol␦ has been shown to be predominantly responsible for lagging DNA strand synthesis during replication (1) but is also required for base and nucleotide excision repair (2)(3)(4), break induced replication and homologous recombination (HR) (5,6). Vertebrate Pol␦ consists of four subunits: POLD1/p125, the catalytic subunit, POLD2/p50, POLD3/p66 and POLD4/p12, which serve regulatory functions (7,8). Consistent with the central role of Pol␦ in genome replication and repair, POLD1 and POLD2 are both essential for cell proliferation (9)(10)(11). However, while the POLD3 ortholog in the fission yeast Schizosaccharomyces pombe is essential (12), the ortholog in the budding yeast Saccharomyces cerevisiae, Pol32, is not (11). Pol32-deficient S. cerevisiae exhibit defects in Poldependent mutagenesis as well as in double-strand break repair by break-induced DNA replication (5)(6)(13)(14)(15). Recent reports have explained the role of Pol32 in mutagenesis by showing that, along with Pol31 (the yeast ortholog of POLD2) it is an integral component of the specialized translesion synthesis (TLS) polymerase Pol (16,17). Moreover, POLD2 and POLD3 are common subunits of both Pol␦ and Pol in mammalian cells (16)(17)(18)(19), interacting with the iron-sulphur clusters found in the C-terminus of both Pol␦ and Pol (19). The current model of TLS suggests that when replicative polymerases encounter DNA lesions they hand off DNA synthesis to one of a number of specialized translesion polymerases, which are able to insert nucleotides opposite the lesion. The resulting mismatch is further extended by Pol , creating a conventional 3 end from which the replicative polymerases can continue DNA synthesis (20,21). It has been suggested that POLD2 and POLD3, being subunits of both Pol␦ and Pol , could contribute to this switch (18) although there is currently no direct evidence for this idea.
The in vivo role of POLD3 in vertebrate replication and TLS has not yet been defined. Here we have generated POLD3-deficient cells (pold3 cells) from the chicken DT40 B lymphocyte line. DT40 cells provide an opportunity for precisely examining in vivo TLS across abasic sites by sequencing the constitutively diversifying immunoglobulin light-chain variable gene (IgL). During avian IgL diversification, the action of activation induced deaminase (AID) leads to conversion of dC to dU, which is then efficiently excised by uracil DNA glycosylase to yield a high frequency of abasic sites (22,23), the most common spontaneouslyarising lesion in the chromosomal DNA (24), within a defined window of around 500 bp. In DT40 these abasic sites can induce gene conversion with a set of homeologous upstream IgL pseudogenes, resulting in templated mutagenesis or can be bypassed directly by TLS resulting in nontemplated mutagenesis at dC/dG basepairs (22,(25)(26)(27)(28). Here we show that pold3 cells replicate with essentially normal kinetics but exhibit a 4.5-fold decrease in the nontemplated immunoglobulin mutagenesis, indicating that POLD3 is required for efficient TLS over abasic sites. Further, pold3 cells exhibit a marked defect in maintaining replication fork progression following UV irradiation. Neither of these phenotypes is observed in cells lacking both Pol and Pol , and combined loss of Pol, Pol and POLD3 results in lethality. Together, these results are consistent with POLD3 playing a critical role in TLS in DT40 cells that is substantially independent of Pol .

Disruption of POLD3 in DT40 cells
Constructs for targeting and disrupting the chicken POLD3 locus were generated from genomic polymerase chain reaction (PCR) products combined with puro R and bsr R selection-marker cassettes ( Figure 1A). Genomic DNA sequences were amplified using primers 5 -GGGGCTCG AGCTGCAGCAAGTGAGTTGGTGTTCCTTC-3 and 5 -GGGGAATTCCTGTGAGCTCTGCTTAGCAGC TGTAGGACC-3 for the 5 -arm of the targeting construct and 5 -GGGACTAGTGTCAACTCAGTGCCTGGGC AGCCAAAGGAGG-3 and 5 -GGGGCGGCCGCCCT TCCTCATCACTGCTGTCAGACTCC-3 for the 3 -arm of the targeting construct. The resulting 3.6 kb 3 -arm and 1.5 kb 5 -arm were cloned into the XhoI-EcoRI and SpeI-NotI sites, respectively, of pBlueScript SK. puro R and bsr R selection cassettes, flanked by loxP sequences, were inserted into the BamHI site to generate pold3-puro R and pold3-bsr R targeting constructs. To generate pold3 cells, wild-type DT40 cells were transfected sequentially with pold3-puro R and pold3-bsr R . Targeted integration of the constructs was detected by Southern blotting of EcoRVdigested genomic DNA and a 0.70 kb probe generated by PCR from the genomic POLD3 locus with the primers 5 -GGCAGCCACGTGGGTCAGGGGCC-3 and 5 -CTGGAGCTGTACACTGCACTGTTGG-3 . Targeting efficiency for the first and second allele was 12.5% (5/40) and 2% (2/100) respectively.

Assessment of sensitivity of cells to genotoxic agents
Sensitivity to ␥ -rays, UV light, cisplatin and camptothecin was measured as a fraction of surviving colonies. A colonyformation assay was performed as described previously (29). Briefly, to analyze sensitivity to ␥ -rays, cells were plated into medium containing methylcellulose and subsequently irradiated with a 137 Cs ␥ -ray source. For exposure of cells to UV light, 3 × 10 5 cells were suspended in 0.5 ml of PBS (phosphate-buffered saline) containing 1% FCS (fetal calf serum) in 6-well plates and irradiated with UVC (254 nm wavelength). For exposure of cells to cisplatin (Nihon-Kayaku, Tokyo, Japan), 1 × 10 5 cells were treated for 1 h at 39.5 • C in 1 ml of complete medium containing cisplatin. Sensitivity to camptothecin, Methyl methansulfonate (MMS) and H 2 O 2 was measured as a fraction of living cells after proliferation in liquid culture. Cells were exposed to camptothecin continuously. For exposure of cells to MMS or H 2 O 2 , 2 × 10 5 cells were treated for 1 h at 39.5 • C in 1 ml of PBS containing 1% FCS and MMS or complete medium containing H 2 O 2 . Two hundred cells were seeded into 384-well white plates (#6007680 Perkin Elmer Life Sciences, Waltham, MA, USA) with 40 l of medium per well. Plates were incubated at 39.5 • C for 72 h. Cell survival was determined using the ATPlite 1-step kit (PerkinElmer). Briefly, 40 l ATPlite solution was added to each well. After 5 min, luminescence was measured by Envision 2104 Multilabel Reader (PerkinElmer).

Flow-cytometric analysis of cell-cycle progression
To analyze cell-cycle progression, 1 × 10 6 cells were exposed for 10 min to 20 M 5-bromo-2'-deoxyuridine (Br-dUrd; Nacalai Tesque, Kyoto, Japan), then harvested and fixed with 70% ethanol. Fixed cells were incubated with 2 M HCl containing Triton X-100 at 0.5%, treated with mouse anti-BrdUrd monoclonal antibody (BD PharMingen, San Diego, CA, USA) and then with FITC-conjugated anti-mouse IgG antibody (Southern Biotechnology Associates, Birmingham, AL, USA). Cells were resuspended in PBS containing propidium iodide (PI) at 5 g/ml for subsequent analysis using FACScaliber (BD Biosciences). To measure the progression of the cell cycle, cells were synchronized at the G 1 phase using centrifugal counterflow elutriation (Hitachi Industrial, Tokyo, Japan). The cell suspension (∼5 × 10 7 cells) was loaded at a flow rate of 11 ml/min into an elutriation chamber running at 2000 rpm. Cell synchrony was confirmed by Fluorescence-activated cell sorting (FACS) analysis.

Dynamic molecular combing and immunofluorescent detection
Asynchronously growing DT40 cells were sequentially labelled for 15 min with 25 M IdU and for 15 min with 25 M CIdU. UV treated cells were irradiated at 20 J/m 2 just before the CldU treatment. At the end of the labeling period (30 min), cells were placed in ice cold 1× PBS (1 volume of cells for 2 volumes of 1× PBS) and centrifuged at 250 g for 5 min at 4 • C, washed in ice-cold PBS, and resuspended in PBS to a final concentration of 1 × 10 6 cells/ml. Three microliter of the cell suspension was spotted onto clean glass Superfrost slides and lysed with 7 l of 0.5% sodium dodecyl sulphate (SDS) in 200 mM Tris-HCl (pH 5.5) and 50 mM ethylenediaminetetraacetic acid (EDTA) (5 min, at room temperature). Slides were tilted at 15 • to horizontal, allowing the DNA to run slowly down the slide. Slides were then air dried and fixed in 3:1 methanol/acetic acid and stored at 4 • C before immunolabeling. IdU, CldU, DNA revelations and analysis were performed as described (30), with minor modifications: the DNA was denatured for 30 min in 2.5 N HCl, and CldU was detected using rat anti BrdU (ABD Serotec, Raleigh, NC, USA) at 1/750. A stretching factor of 2.6 for conversion from m to kb was applied, as previously described for the method in (31). Slides were mounted in 10% 1× PBS and 90% glycerol, kept at −20 • C and imaged using a Nikon C1-si confocal microscope.

AID overexpression by retrovirus infection
Wild-type and pold3 cells were inoculated in 96-well plates to obtain single colonies (Supplementary Figure S4). Several single colonies were picked up and genomic DNA was extracted. After the DNA sequence of the V(D)J locus was analyzed, clones without any mutation or gene conversion in the V(D)J locus were obtained. AID overexpression was carried out by infection of retrovirus containing the AID gene and the internal ribosomal entry site (IRES) followed by the Green fluorescent protein (GFP) gene (Supplementary Figure S4), as described previously (27,32). The efficiency of infection was about 70%, as assayed by GFP ex-pression. The per division rate of Ig-gene conversion and hypermutation (events/IgV segment/division) was calculated assuming doubling times of 7.6 and 11 h for wild-type and pold3 cells respectively.

Analysis of Ig V diversification
Genomic DNA was extracted at 14 days post infection. Using primers 5 -CAGGAGCTCGCGGGGCCGTCACTG ATTGCCG-3 , in the V leader intron and 5 -GCGCAA GCTTCCCCAGCCTGCCGCCAAGTCCAAG-3 , in the J-C intron, the rearranged V segments were PCR amplified, with Prime Star DNA polymerase (Takarabio, Shiga, Japan), a high fidelity thermostable polymerase, cloned into pTOPO-Zeroblunt (Invitrogen) and sequenced with the M13 forward (−20) primer. Sequence alignment with DNASIS-MAC v3.3 (Hitachi Solutions, Tokyo, Japan) allowed identification of changes from the parental sequences in each clone. The method for classification of gene conversion and non-templated point mutations was as previously described (33,34,35).

Creation of the anti-POLD3 antibody
Chicken POLD3 was amplified from DT40 cDNA with the primers 5 -ATAGGATCCATGGAGGACGAGCT GTAC-3 and 5 -GGCGAATTCTCATTTCTTCTGAC AGAAG-3 and cloned into pGEX-6P1 (Amersham, Buckinghamshire, UK) for expression as in GST-tagged form in Escherichia coli [BL21 STAR (DE3) (Invitrogen)]. Cells were induced with 1 mM Isopropyl ␤-D-1thiogalactopyranoside (IPTG) and cultured for a further 4 h at 22 • C. Cells were harvested, disrupted by sonication and the lysate cleared by ultracentrifugation. The lysate was passed over a glutathione sepharose 4B column and the GST-POLD3 protein eluted with 50 mM Tris-HCL pH 8.0, 150 mM NaCl, 20 mM glutathione. The eluted protein was then applied to a 1 ml HiTrap Q XL column (Amersham) and eluted with a NaCl gradient. GST-POLD3 eluted at 750 mM NaCl. For antibody production, two rabbits were immunized (Eurogentec, Seraing, Belgium) with 200 g GST-POLD3. Serum was collected 4 weeks after the final immunization. The antibody was purified against a his-tagged POLD3 immobilized on an agarose column.

Protein purification and primer extension assays
The human Pol␦ holoenzyme with or without POLD3, with N-terminal His-tagged p50, was expressed using a baculovirus vector (pBacPAK9, Clontech, Palo Alto, CA, USA) in insect cells (High Five, Life Technologies, Palo Alto, CA, USA), as described previously (36). The concentrations and purity of the proteins was estimated from the intensity of protein bands in an SDS-polyacrylamide gel (Supplementary Figure S6). For primer-extension analysis, DNA synthesis was carried out with 0.06 pmol 32 P-labeled M17 primer (AGC-TATGACCATGATTA) annealed with a 49-mer template oligo DNA (AGCTACCATGCCTGCACGAATXAAG-CAATTCGTAATCATGGTCATAGCT), where X can be an abasic site. The assay was carried out in a reaction mixture (5 l) containing 30 mM HEPES-NaOH (pH 7.4), 7 mM MgCl 2 , 8 mM NaCl, 0.5 mM dithiothreitol and 10 M each dNTP in the presence of 2 nM of Pol␦ and 50 nM of PCNA for 15 min at 37 • C. Used dNTP concentration reflects that in human cycling cells (37). At the end of the reaction, the products were denatured with formamide and loaded onto 15.6% polyacrylamide gels containing 7 M urea in Tris, Boric acid, EDTA (TBE) buffer (89 mM Tris, 89 mM boric acid, 2 mM EDTA). After electrophoresis, radioactivity was measured with a Fuji Image analyzer, FLA2500 (Fujifilm, Tokyo, Japan).

DT40 cells lacking POLD3 are viable but exhibit a moderate S phase delay
To explore the role of POLD3, we disrupted the POLD3 gene in the chicken DT40 B cell line, creating pold3 cells ( Figure 1A and B). Loss of POLD3 expression was verified by RT-PCR and western blot analysis ( Figure 1C and D). pold3 cells are able to proliferate although their doubling time was 1.5× longer than that of wild-type cells (Figure 1E) due to a delay in late S-phase and a higher frequency of apoptosis (the left-most R5 gate, Figure 1F). To confirm this, we monitored cell cycle progression after release from synchronization in G 1 . Wild-type and pold3 cells completed S-phase at 6 and 9 h after release respectively and returned to the next G 1 -phase at 9 and 12 h ( Figure 1G). Thus, completion of DNA replication is delayed in the absence of POLD3.

pold3 cells have a normal velocity of DNA synthesis but an increase in unidirectional fork movement
To understand the nature of the S phase delay, we examined the kinetics of unperturbed DNA replication using molecular combing (Tables 1 and 2). Strikingly, the replication rate for wild-type and pold3 cells in unperturbed conditions was not significantly different at 1.36 ± 0.05 kbp/min and 1.28 ± 0.03 kbp/min respectively (P > 0.34). The average inter-origin distance was also comparable between the two lines. These findings indicate that loss of POLD3 does not attenuate global progression of DNA replication forks on undamaged DNA templates (Table 1). However, pold3 cells did exhibit an increase (20% as opposed to 13% in wild-type cells) in the proportion of unidirectional forks, in which one of the two forks heading away from a presumed origin is missing (Table 2, Supplementary Figure S1). Thus, POLD3 might be required for releasing replication forks that stall at spontaneously arising DNA damage on genomic DNA. pold3 cells exhibit sensitivity to a wide range of DNA damaging agents pold3 cells exhibited sensitivity to a broad range of DNA damaging agents, including ionizing radiation (␥ -rays), UV irradiation, the methylatingagent (methyl methane sulphonate, MMS), H 2 O 2 and the chemical crosslinker (cisplatin) (Figure 2). This hypersensitivity to DNA damage was reversed by the ectopic expression of chicken POLD3 (Figure 2).
Since DNA polymerase ␦ and are involved in a number of DNA repair and damage tolerance pathways (38), we next sought to define the pathways affected by loss of POLD3. The sensitivity of cells to the methylating agent, MMS and oxidising agent, H 2 O 2 prompted us to test whether pold3 cells exhibit a defect in base excision repair. However, gap filling during base-excision repair (39) occurred normally in pold3 cells (Supplementary Figure S2). To investigate the involvement of POLD3 in nucleotide excision repair, we performed epistasis analysis of POLD3 and XPA, an essential protein of nucleotide excision repair (40). Inactivation of the XPA and POLD3 genes had a synergistic impact on cellular sensitivity to UV, suggesting that POLD3 contributes to cellular tolerance of UV independently of nucleotide excision repair (Supplementary Figure  S3). To investigate the involvement of POLD3 in HR, we measured cellular sensitivity to the topoisomerase I poison, camptothecin and gene targeting efficiency. pold3 cells are not hypersensitive to camptothecin ( Figure 2) and are proficient in gene targeting (Supplementary Table S1). Taken together, hypersensitivity of the pold3 cells to various DNA damaging agents is not attributable to defects in the excision repair pathways or in HR.

POLD3 is required for efficient Ig V gene diversification by TLS
The wide-ranging hypersensitivity of pold3 cells to DNAdamaging agents is reminiscent of the phenotype of TLSdeficient cells (41)(42)(43). To examine the role of POLD3 in TLS, we took advantage of the diversification of the Ig V light chain locus of DT40 cells, described in the Introduction and summarized in Supplementary Figures S4 and S5.
To examine Ig V diversification, we ectopically overexpressed AID and isolated at least three subclones each of wild-type and pold3 cells (Supplementary Figure S4). Following two weeks clonal expansion, we compared the frequency and spectrum of mutations in the IgL variable segment in wild-type and pold3 clones (Supplementary Figure  S4). pold3 cells exhibited a 4.5-fold reduction in the rate of non-templated mutations ( Figure 3A and B). This reduced frequency of TLS was accompanied by a two-fold increase in the frequency of Ig gene conversion events suggesting that the failure of POLD3-dependent TLS results in more lesions being available to trigger Ig gene conversion and showing that POLD3 is not required for robust intragenic HR. We next looked in detail at the mutation spectrum of abasic site bypass. Under conditions of AID overexpression, 62% of the TLS in wild-type cells involves insertion of dC opposite the abasic sites created on both strands, leading to dC to dG transversions. This reaction is largely dependent on the deoxycytidyl transferase activity of REV1 (44,45). Following disruption of POLD3, the frequency of mutagenesis is decreased but 78% of the remaining mutations indicate insertion of dC opposite the abasic site. Intriguingly, few remaining mutations at dC in the pold3 mutant appear to be highly strand biased. However, confirming this result will require a very much more substantial dataset.
Cells lacking Pol alone do not tolerate AID overexpression, dying within a few days of retroviral infection (20). However, cells deficient in both Pol and Pol are able to Profile of the DNA replication kinetics in indicated cells with or without UV irradiation was analysed by molecular combing.  tolerate AID overexpression and exhibit a level of gene conversion and point mutagenesis comparable to wild-type cells, but with a marked increase in dA incorporation opposite the AID-induced abasic sites (20) (Figure 3). Thus, the immunoglobulin mutagenesis in cells lacking POLD3 and those lacking both Pol and Pol is quite distinct. This suggests that while POLD3 contributes to TLS in the immunoglobulin loci, it does so substantially independently of Pol . Moreover, polζ cells displayed significantly higher sensitivity to the ectopic AID expression in comparison with pold3 cells, further suggesting functional independence of POLD3 and Pol .

Cells deficient in POLD3, Pol and Pol are inviable
To further test whether POLD3 acts independently of the Pol/Pol TLS axis, we conditionally disrupted POLD3 in polη/polζ double mutant cells by covering the deletion of POLD3 with a chicken POLD3 transgene, the expression of which was under a tetracycline-repressible promoter (Figure 4A). polη/polζ /pold3 triple mutant cells expressing the cPOLD3 transgene proliferated slightly more slowly than polη/polζ /pold3 +/− cells ( Figure 4B). Upon addition of doxycycline, POLD3 expression was significantly reduced by 24 h and lost completely by 48 h. By 3-4 days the cells had stopped proliferating. This observation further supports the idea that POLD3 plays a role that is independent of Pol and Pol .

POLD3 is required to maintain replication fork progression on UV damaged DNA but not to promote the filling of postreplicative gaps
Previous work in DT40 cells has revealed temporally separated modes of lesion bypass. One mechanism is responsible for timely filling of postreplicative gaps at UV-damage sites, while another operates at or very close to stalled replication forks and maintains normal fork progression on UVdamaged DNA (46). These modes of damage bypass can be assessed by alkaline sucrose gradient sedimentation of newly replicated DNA ( Figure 5A) and DNA molecular combing (Tables 1 and 2, and Figure 5B and C), respectively.
We first asked whether POLD3 is required for postreplicative gap filling. Without exposure to UV, the 14 C thymidine label was incorporated into higher molecular weight DNA similarly in wild-type and pold3 cells (Figure 5A upper panel). This is consistent with the observation that POLD3 is not essential for unperturbed DNA replication (Table 1). However, pold3 cells are also able to reconstitute high molecular weight DNA during a 3-h chase period following UV irradiation ( Figure 5A lower panel) with similar kinetics to wild-type cells. Similarly, polζ cells exhibited reconstitution of higher molecular weight DNA following UV irradiation that was indistinguishable from wild-type cells (42). These data contrast with the defective postreplicative gap filling seen in polη, rad18 and pcna K164R/K164R DT40 cells (46). Thus, neither Pol nor POLD3 play a critical role in postreplicative gap filling at UV damaged sites in DT40.
We next examined replication fork progression after UV irradiation. To do this, we labeled nascent strands with IdU for 15 min, UV irradiated the cells and then continued labeling the nascent strands with CldU ( Figure 5B and Table 1). After DNA combing, we revealed the tracts of CldU and IdU and calculated the ratio between them to allow comparison of the total DNA synthesized before and after UV exposure on a fork-by-fork basis. We plotted the data as a cumulative percentage of forks at each ratio ( Figure 5C). polζ cells exhibit a response indistinguishable from wildtype i.e. a modest reduction in the average length of DNA synthesized during the second labeling period following UV ( Figure 5C). This is consistent with a previous report examining Polζ -deficient mouse cells (47). In contrast, pold3 cells exhibit a significant reduction in the length of DNA synthesized during the second labeling period after UV, consistent with a significantly increased proportion of forks stalling at UV lesions (P = 2.77 × 10 −7 , Supplementary Figure S6). In the time course of this experiment (15 after 20 J/m 2 UV), we estimate that forks would have about a 60% chance of encountering a UV dimer (based on (46), Supplementary Figure S1). Thus, POLD3 is required to maintain replication fork progression on a UV damaged template independently of Polζ . Taken together, these data indicate that POLD3 facilitates TLS independently of Polζ , leading us to speculate that it may facilitate TLS by Pol␦ itself.

POLD3 facilitates invitro TLS past abasic sites by Pol␦
To address the possibility that TLS by Pol␦ might be modulated by POLD3, we purified the intact human Pol␦ holoenzyme (Pol␦ (POLD3 + )) and POLD3-deficient holoenzyme (Pol␦ (POLD3 − )). We employed the human Pol␦ holoenzyme since the method for protein purification has been established (48). Pol␦ (POLD3 + ) and Pol␦ (POLD3 − ) were expressed and purified with the same efficiency (Supplementary Figure S7), indicating that the absence of POLD3 does not diminish the stability of the other three components of the holoenzyme. While the Proliferating cell nuclear antigen (PCNA) binding capability of POLD3 may increase the processivity of Pol␦ on a long circular template (49), to focus our analysis on the TLS capability of Pol␦, we used a short (49-mer) linear template. As expected from the in vivo data shown in Table 1 and Figure 5A (upper panels), DNA synthesis by purified Pol␦ (POLD3 + ) and Pol␦ (POLD3 − ) on an intact template DNA strand was comparable (Figure 6A and C). However, bypass replication across an abasic site was significantly reduced in Pol␦ (POLD3 − ) (Figure 6B and C). The bypass is performed in two steps, the insertion of a nucleotide opposite to the lesion followed by extension of the resulting mismatch. Pol␦ (POLD3 + ) and Pol␦ (POLD3 − ) performed the insertion step with comparable efficiency, as the yield of product indicating an incorporation opposite the abasic site ('0' in Figure 6B) relative to that reflecting stalling at the −1 position was very similar with both Pol␦ (POLD3 − ) and Pol␦ (POLD3 + ). On the other hand, Pol␦ (POLD3 + ) was able to extend beyond the abasic site and complete DNA synthesis much more efficiently than Pol␦ (POLD3 − ) ( Figure 6B and C). The reduced extension by Pol␦ (POLD3 − ) might be attributable to (i) higher exonuclease activity of Pol␦ (POLD3 − ) and/or to (ii) lower extension ability from the mismatched primer terminus than Pol␦ (POLD3 + ). To address these possibilities, we measured degradation and extension using a 28-mer primer, which carries dA in its 3 terminus opposite the abasic site in template DNA. The ratio between intact and degraded primers in Pol␦ (POLD3 + ) and Pol␦ (POLD3 − ) was indistinguishable ( Figure 6D). Moreover, the kinetics of the primer degradation in the absence of deoxynucleotides was also indistinguishable between Pol␦ (POLD3 + ) and Pol␦ (POLD3 − ) ( Figure 6E). Thus, we propose that POLD3 is required for effective coupling of nucleotide insertion with subsequent extension thereby enhancing the overall processivity of Pol␦ during abasic site bypass.

DISCUSSION
Here we have shown that a vertebrate cell line is capable of proliferating and replicating undamaged DNA with apparently normal kinetics in the absence of the POLD3 subunit of Pol␦. This suggests that POLD3 does not play a major role in the central function of Pol␦ as the main lagging strand replicase (1). This rather surprising finding contrasts with the essential function of the POLD3/Pol32 ortholog of the fission yeast S. pombe (12). Further, while Pol␦ is involved in excision repair (2)(3)(4) and recombination (5-6,50), we could not detect any gross defects in these repair pathways in pold3 cells. Rather, we found that POLD3 contributes significantly to cellular tolerance of a wide variety of DNA damaging agents by presumably promoting TLS. It has been known for some time that Pol32 plays a role in TLS in S. cerevisiae and that genetically it played roles that are both dependent and independent of REV3, the catalytic subunit of Pol , depending on the mutator examined (13)(14)(15)(51)(52)(53). Recent biochemical studies have shown that Pol32 is not only a subunit of Pol␦ but is also an integral component of Pol (16,17). This has led to the attractive suggestion that this association with Pol explains the role of POLD3 in TLS and has led to the suggestion that Pol32 may facilitate switching between Pol␦ and Pol (16)(17)(18)(19). POLD3 and POLD2 are also subunits of human Pol , interacting with the C terminal iron-sulphur clusters of REV3 in a similar manner to that reported in yeast (18,19). However, to date genetic data addressing the role played by vertebrate POLD3 in TLS have been lacking. The data we present here highlight striking phenotypic differences between cells lacking POLD3 and those lacking Pol . Together with the lethality that results when both are disrupted the data suggest that the main role played by POLD3 in TLS in DT40 cells is independent of Pol .
An important question is how POLD3 promotes TLS independently of Pol . An obvious and attractive potential explanation is that POLD3 facilitates lesion bypass by Pol␦ itself. While Pol␦ is stalled by many lesions, there is good in vitro evidence that it is able to insert and extend over a number of single base lesions both with and without PCNA (48,(54)(55)(56), including abasic sites (48,(56)(57). We show that purified human Pol␦ (POLD3 + ) is able to insert and extend over an abasic site at a physiological concentration of dNTP ( Figure 6). Pol␦ preferentially inserts dA (57), an activity that could explain the bias toward the incorporation of dA we observed in the IgV gene in the absence of Pol and Pol. How might POLD3 facilitate TLS by Pol␦, given that it appears largely dispensable for general replication? We show that while Pol␦ (POLD3 + ) and Pol␦ (POLD3 − ) insert a single nucleotide opposite an abasic site with very similar efficiency, Pol␦ (POLD3 + ) is much more proficient at extension of the resulting mismatch ( Figure 6B and C). This more efficient extension is attributable to higher processivity, but not lower proofreading exonuclease activity, in the presence of POLD3 ( Figure 6D and E   that Pol␦ (POLD3 + ) would cycle opposite the abasic site with nucleotide insertion balanced by exonucleolytic excision, with occasional extension events from the inserted nucleotide allowing for completion of the bypass of the abasic site. A possible explanation for the role of POLD3 would be that it helps prevent Pol␦ dissociating from the mismatched primer terminus, increasing the chance of a successful extension by POLD1. Importantly, our data do not exclude cooperation between POLD3 and Pol . Despite our data supporting Pol␦ being able to perform abasic site TLS on its own, independently of Pol , it may remain the case that the efficiency of extension in vivo is facilitated by Pol , which is able to complete bypass regardless of which polymerase mediated the insertion step. However, this role may become essential when the ability of Pol␦ to extend directly is compromised by loss of POLD3. Such a model would explain the synthetic lethality of cells lacking both POLD3 and Pol and would suggest that the non-redundant function of the two proteins is in mediating the extension step of abasic site bypass.
Our study also indicates that POLD3 contributes to TLS operating at or close replication forks but not obviously to post-replicative gap filling. Previous studies both in DT40 cells (46) and mice (47) have shown that REV1 is also important for maintaining fork progression on damaged DNA and for effective Ig hypermutation (44,58). It is thus tempting to speculate that the two proteins may cooperate. In-deed, a physical interaction between Pol32 and REV1 has been described in S. cerevisiae (59), although this interaction has not been demonstrated in vertebrates. Nonetheless, it will be interesting to further explore the genetic and physical relationship between REV1 and POLD3 in vertebrates to examine whether POLD3 and REV1 collaborate to facilitate restart of replication by Pol␦.