DNA Polymerase δ Is Required for Early Mammalian Embryogenesis

Background In eukaryotic cells, DNA polymerase δ (Polδ), whose catalytic subunit p125 is encoded in the Pold1 gene, plays a central role in chromosomal DNA replication, repair, and recombination. However, the physiological role of the Polδ in mammalian development has not been thoroughly investigated. Methodology/Principal Findings To examine this role, we used a gene targeting strategy to generate two kinds of Pold1 mutant mice: Polδ-null (Pold1 −/−) mice and D400A exchanged Polδ (Pold1 exo/exo) mice. The D400A exchange caused deficient 3′–5′ exonuclease activity in the Polδ protein. In Polδ-null mice, heterozygous mice developed normally despite a reduction in Pold1 protein quantity. In contrast, homozygous Pold1 −/− mice suffered from peri-implantation lethality. Although Pold1 −/− blastocysts appeared normal, their in vitro culture showed defects in outgrowth proliferation and DNA synthesis and frequent spontaneous apoptosis, indicating Polδ participates in DNA replication during mouse embryogenesis. In Pold1 exo/exo mice, although heterozygous Pold1 exo/+ mice were normal and healthy, Pold1 exo/exo and Pold1 exo/− mice suffered from tumorigenesis. Conclusions These results clearly demonstrate that DNA polymerase δ is essential for mammalian early embryogenesis and that the 3′–5′ exonuclease activity of DNA polymerase δ is dispensable for normal development but necessary to suppress tumorigenesis.


Introduction
DNA replication is the process for transmitting genetic information to future cells and offspring. To date, at least 14 eukaryotic DNA-dependent DNA polymerases have been found [1][2][3][4]. DNA polymerase a (Pola), DNA polymerase d (Pold) and DNA polymerase e (Pole) are thought to be distinct DNA polymerases that directly participate in DNA replication [3,5]. These three polymerases belong to the type B DNA polymerase subfamily [6]. Pola forms a Pola/RNA primase complex that synthesizes short RNA-DNA primers for DNA synthesis. On the other hand, Pold and Pole are believed to extend these short primers; Pold is mainly responsible for lagging strand synthesis while Pole is thought to be responsible for leading strand synthesis (reviewed in [3,7]). However, the precise molecular roles of Pold and Pole remain unclear. Furthermore, in addition to DNA replication, it is known that these elongating polymerases are also involved in several DNA repair and recombination pathways [1,5].
Our interest is in Pold, which is a highly processive polymerase that associates with PCNA. The Pold holoenzyme consists of several subunits: the catalytic subunit p125 (Pold1), which is encoded in the Pold1 gene and is highly conserved among eukaryotes; and two to four more divergent smaller subunits which are believed to regulate PCNA binding, Pold function and its structure. Pold1 comprises of two primary functional domains: an exonuclease domain near the N-terminus that catalyzes 39-59 exonucleolytic proofreading and the subsequent polymerase domain that catalyzes DNA synthesis.
The molecular functions of Pold had been revealed mainly by using biochemical analysis and yeast genetics. Studies using the complete replication of DNA plasmids containing the Simian Virus 40 (SV40) DNA replication origin with purified mammalian cell extracts showed that Pold is required for in vitro DNA replication [8], while a study using Xenopus egg extracts immunodepleted of Pold also showed the indispensability of Pold for DNA replication [9]. In fission yeast, temperature-sensitive mutants and tetrad analysis from diploid heterozygous Pold1disrupted mutants showed that Pold1 was essential for viability while disruption of Pold1 resulted in a terminal phenotype similar to a cell division cycle mutant, indicating that Pold was involved in DNA replication [4]. The function of each Pold1 domain has been investigated extensively by using various yeast mutants. For example, mutations disrupting 39-59 exonuclease activity caused a strong mutator phenotype in budding yeast [10,11]. Mutations in motif A of the polymerase domain also expressed a mutator phenotype and hypersensitivity to hydroxyurea and methyl-methane sulfonate in budding yeast [12,13]. Zinc finger domain deletion mutants showed that the zinc finger domain of Pold1 is responsible for binding to the regulatory subunit [14]. However, despite many studies revealing the molecular roles of Pold, the necessity of Pold in mammalian cell has not been uncovered by a genetic approach.
Compared to the molecular functions of Pold in eukaryotic cells, the physiological role of entire Pold in higher eukaryotes remains obscure. For example, Zebrafish Pold1 disrupted mutants were found as a result of positional cloning of the flathead gene [15]. Flathead mutants display specific defects in late proliferative zones, such as eyes, brain and cartilaginous elements of the visceral head skeleton, where cells showed compromised DNA replication. However, the zebrafish mutants did develop normally during their early stages. This may be the result of a functionally redundant gene of the Pold1 gene due to genomic duplication during teleost evolution. Using mouse genetics, mice deficient in Pold 39-59 exonucelase activity showed that 39-59 exonuclease activity suppresses the risk of tumorigenesis as a result of decreasing the spontaneous mutation rate [16,17]. Mice harboring a mutation in motif A of the polymerase domain were embryonic lethal in homozygotes and experienced genomic instability in heterozygotes as compared to wild types. One motif A mutation, L604K, reduced the life span and accelerated tumorigenesis in heterozygous mice [18]. However, when, how and why these motif A mutant homozygotes suffered from death during embryonic development has not been elucidated well.
In our study, to uncover the physiological role of Pold in mammalian development more clearly, we generated mice lacking functional Pold by gene targeting methods in a Pold1 2 allele and examining its effects. For a more comprehensive analysis, we generated mice lacking 39-59 exonuclease activity in the Pold of a Pold1 exo allele. Here, we report that peri-implantation lethality accompanies the loss of functional Pold. We speculate this is due to an impairment in cell proliferation, a defect in DNA synthesis and an occurrence of apoptosis observed in an in vitro blastocyst outgrowth. Pold is the first DNA polymerase to be studied in such a model offering new insight on how DNA polymerases are involved in nuclear chromosomal replication.

Generation of Pold1 gene targeted mice
To analyze the physiological function of the entire Pold and its 39-59 exonuclease activity, we created two types of Pold1 mutant mice. The Pold1 gene encodes the catalytic subunit p125 of the Pold complex. This subunit contains the DNA polymerase domain and 39-59 exonuclease domain. To disrupt the functional murine Pold1 protein, we inserted a splice/polyadenylation (poly(A)) signal derived from SV40 into the upstream region encoding the Pold polymerase domain (schematic in Fig. 1A). Using a point mutation of GCC from GAC (wild-type), 39-59 exonuclease activity deficient mice were created by D400A exchange, which is located in the exonuclease active site (ExoII) and with which mice suffered from higher tumor susceptibility [16,17]. D400A mutation in mice has been confirmed to cause inhibition of 39-59 exonuclease activity by in vitro biochemical analysis [17]. We used the Cremediated loxP recombination system in our gene targeting strategy to generate the two kinds of Pold1 mutant mice: Poldnull (Pold1 2/2 ) mice and its 39-59 exonuclease activity deficient (Pold1 exo/exo ) mice (Fig. 1B) [19].
The Pold1 gene targeting vector was electroporated into C57BL/6J embryonic stem (ES) cells. G418-resistant clones were screened for homologous recombination by PCR (Fig. 1C) and Southern blotting. A point mutation (D400A) was verified by DNA sequencing. The targeted ES cell clones were microinjected into Balb/c blastocysts to generate chimeric mice which transmitted the targeted allele through the germ line. Chimeras were crossed with C57BL/6J females, yielding Pold1 +/2 mice. The Pold 39-59 exonuclease activity deficient mice were generated by crossing Pold1 +/2 mice and Cre expressed transgenic animals which contained a murine Sycp1 gene promoter-Cre transgene that expressed Cre-recombinase in germ-line cells. Both Pold1 +/2 and Pold1 exo/+ mice were identified by PCR and Southern blot analysis ( Fig. 1D and 1E).
To confirm the expression of the Pold1 gene from these targeted alleles, we analyzed mRNA from targeted ES cells. The Pold1 exo/+ ES cells were obtained by electroporating circular Cre-pac plasmids into Pold1 +/2 ES cells and selecting adequate clones [20]. RT-PCR was performed on transcripts from Pold1 +/2 and Pold1 exo/+ ES cells using two primers that covered the full-length Pold1 coding region. Direct sequencing of these RT-PCR products showed that substituted alleles (GACRGCC) were expressed in Pold1 exo/+ ES cells, but not in Pold1 +/2 ES cells (Fig. 1F). These results revealed that full-length mRNA was not expressed from the Pold1 2 allele. To investigate the expression from mutated alleles in detail, we determined transcriptional end sites from Pold1 2 alleles by 39 RACE analysis. In 39RACE products from Pold1 +/2 ES cells, both full-length Pold1 coding fragments and truncated fragments were observed; while in the products from Pold1 exo/+ ES cells, only the full-length fragment was observed. After subcloning these fragments, DNA sequencing was performed. DNA sequencing showed that truncated fragments from Pold1 +/2 ES cells contained a substituted sequence (GCC), the 39 terminus was prematurely stopped by the inserted SV40 poly(A) signal sequence, and these truncated fragments consisted of two aberrant splicing forms (Fig. 1G). The full-length fragment sequences from Pold1 +/2 ES cells only contained the wild-type sequence (GAC) and used an endogenous poly(A) signal sequence from the Pold1 gene. In Pold1 exo/+ ES cells, mRNA terminated normally and included both substituted and wild-type sequences (two and three of five subclones, respectively). These results suggest that mRNA terminated by the inserted SV40 poly(A) signal was expressed by the Pold1 2 allele and the full-length transcript with the D400A (GACRGCC) substitution was expressed by the Pold1 exo allele. To confirm that Pold1 2 allele was impaired in the expression of the Pold1 protein, we analyzed the Pold1 protein by Western blotting. Western blotting of the extracts prepared from Pold1 +/+ and Pold1 +/2 E12.5 whole embryos with an anti-Pold1 C-terminal region antibody showed that the Pold1 protein quantity was reduced to about half in Pold1 +/2 embryos, while the Pold1 protein quantity from Pold1 +/+ and Pold1 exo/exo E12.5 embryonic extracts were comparable (Fig. 1H). These results suggest that the Pold1 2 allele did not function properly due to the prematurely terminated polymerase domain and that the Pold1 exo allele expressed quantities of Pold similar to the wild-type allele despite lacking 39-59 exonuclease activity.

Deficiency of Pold1 causes embryonic lethal around periimplantation
Heterozygous Pold1 +/2 mice were apparently normal and fertile with no detectable developmental abnormalities over 18 months. No defect in the proliferation of embryonic fibroblasts from these animals was found (data not shown). In contrast, no homozygous Pold1 2/2 mice were detected among the 256 live births from Pold1 +/2 intercrosses (Table 1), indicating that one functional Pold1 allele is sufficient for embryonic and postnatal development, whereas inactivation of both alleles leads to embryonic lethality. To determine the time of death during the development of Pold1 2/2 embryos, embryos were collected from intercrosses of Pold1 +/2 mice at different times of gestation and individual embryos were genotyped by PCR. No homozygous Pold1-deficient embryos were found at E7.5 or beyond (Table 1). We also collected embryos at day E3.5 and E4.5 by flushing the uteri of pregnant females. The homozygous Pold1 2/2 mutant blastocysts were morphologically indistinguishable from wild-type and heterozygous embryos (Fig. 2A). These observations indicate that Pold1 2/2 embryos died between E4.5 and E7.5.

Pold1 is required for the proliferation of blastocyst outgrowth
To further characterize the developmental abnormality of Pold1-deficient embryos, we collected E3.5 blastocysts derived from Pold1 +/2 intercrosses, cultured them in vitro individually for several days, took photographs, and subsequently genotyped them by PCR. In Pold1 +/+ and Pold1 +/2 blastocysts, trophoblasts spread over the culture dish after hatching from the zona pellucida while an inner cell mass (ICM) grew on the trophoblast sheet after 3 days of culture. In Pold1 2/2 embryos, the trophoblasts again attached and spread over the dish, but the trophoblast sheet spread slowly and the ICM failed to proliferate and degenerated soon after appearing (Fig. 2B). Pold1-null blastocysts cultured in vitro for a day, which roughly corresponded to E4.5 in vivo, had no apparent difference from wild-type and heterozygous embryos. Pold1 2/2 blastocysts cultured for 3 days, roughly corresponding to E6.5, uncovered a serious proliferation defect.

Lack of Pold1 causes DNA synthesis defects
To uncover whether the proliferation defect of the cultured Pold1 2/2 blastocysts is due to DNA synthesis inhibition caused by Pold disruption, we assessed DNA synthesis in cultured blastocysts by the bromodeoxyuridine (BrdU) incorporation assay. We cultured blastocysts for 1 or 3 days in normal ES medium and then cultured them in the presence of BrdU for 3 hr. These cultured blastocysts were immunostained with a specific antibody for BrdU. In addition, to assess whether embryonic cells in the cultured Pold1 2/2 blastocysts entered the mitotic phase, we simultaneously immunostained them with an antibody against phosphohistone H3 (Ser-10), a common mitotic marker used to mark M-phase cells [21]. Incorporation of BrdU into 1 day cultured blastocyst nuclei revealed no differences between each genotype, as all embryos showed extensive labeling (Fig. 3A). We also found phosphohistone H3-positive cells in all embryos regardless of genotype (Fig. 3A). These results suggested that Pold1 2/2 embryos were able to synthesize new DNA strands, equivalent to the S-phase of the cell cycle, and entered the mitotic phase 1 day after the cultured blastocyst stage. In contrast, in 3 days of cultured blastocyst outgrowth, both the Pold1 +/+ and Pold1 +/2 outgrowths showed strong incorporation of BrdU and frequent mitotic stainings into the ICM and trophoblast giant (TG) cells, but Pold1 2/2 outgrowths showed very little BrdU incorporation and few mitotic cells (Fig. 3B). In TG cells, it has been shown that DNA synthesis occurs by endoreduplication, whereby successive rounds of G and S phases proceed without mitosis [22,23]. As a result, endocycling giant cells acquire massive quantities of DNA in their nuclei, thereby becoming polyploid [24]. This deficiency in Pold1 2/2 outgrowths indicates that the disruption of Pold caused the distinct DNA synthesis defect including the TG cell endoreduplication defect.

Deficiency of Pold1 protein results in spontaneous apoptosis
In the 3rd day of cultured Pold1 2/2 outgrowths stained with DAPI, we found condensed and fragmented micronuclei, possibly indicating apoptotic cell death (Fig. 3B). These micronuclei were rarely found in heterozygous or wild-type outgrowths. To confirm this apoptotic possibility, TUNEL assays were performed on blastocyst outgrowths cultured for 3 days. These aberrant nuclei in Pold1 2/2 outgrowths were all TUNEL-positive, whereas TUNELpositive cells were rare in Pold1 +/+ and Pold1 +/2 outgrowths (Fig. 4). We further explored apoptotic cell death in harvested E3.5 blastocysts. No apparent difference was observed between the Pold1 2/2 blastocysts (n = 4) and the blastocysts with other genotypes (n = 17) (date not shown). These data indicate that proliferation defects in Pold1 2/2 outgrowths were, at least in part, due to an occurrence of apoptosis around hatching.
One 39-59 exonuclease deficient Pold1 exo allele enables embryonic development, but increases tumor susceptibility In Pold 39-59 exonuclease activity deficient mice, it has been shown that the mutant allele does not affect embryo viability and that homozygous mutant mice are fertile. However, recessive mutants are more prone to cancer and death [16,17]. To confirm the physiological role of Pold 39-59 exonuclease activity and to examine an interaction between the Pold1 exo allele and the Pold1 2 allele, we analyzed this exonuclease activity disrupted mice. We found similar observations as those in [16,17] for Pold1 exo/exo and Pold1 exo/+ mice, as both were able to grow normally and were fertile ( Table 2). At embryonic stage E12.5, no morphological differences between Pold1 exo/exo and wild-type embryos were found. Moreover Pold1 exo/2 mice, which were generated by crossing between Pold1 exo/exo and Pold1 +/2 mice, could give birth and were also apparently normal and fertile ( Table 2). These results showed that for mouse development, Pold 39-59 exonuclease activity is dispensable and one Pold1 exo allele is sufficient. Although Pold1 exo/exo and Pold1 exo/2 mice were developmentally normal, they frequently died with abnormally swollen thymuses between 3 and 8 months (9 out of 28 Pold1 exo/exo animals and 5 out of 14 Pold1 exo/2 animals) (Fig. 5A), and many of the surviving mice without swollen thymuses developed nodules on their tails after 12 months (Fig. 5B). These observations are consistent with typical cancer symptoms owing to the lack of Pold 39-59 exonuclease activity [16,17]. Histological analysis of swollen thymuses showed a homogenous population of cells with a high nuclear to cytoplasmic ratio and frequent mitoses, typical morphological features of lymphoblastic lymphoma (date not shown). Pold1 exo/+ mice and wild-type mice did not express such phenotypes on their thymus and tails (more than 20 animals observed for each genotype). Thus, the lack of Pold 39-59 exonuclease activity also increased cancer susceptibility in C57BL/6 mice like as those mixed with C57BL/6J and 129/SvJ as previously seen [17]. Furthermore, we found that Pold1 exo/2 mice developed similar tumors.

Discussion
So far the function of Pold had been mainly studied by biochemical approaches and yeast genetics. This present study revealed the function of Pold by examining its Pold1 knock out mice. Although many DNA polymerases have been clarified, the physiological role of three DNA polymerases directly involved in   chromosomal replication, Pola, Pold and Pole, have not been assessed in Pol knockout mice. In this study, we found that Pold disrupted mice suffered from embryonic lethality after E4.5. To date, three DNA polymerases, Polb [25,26], Polc [27] and Polf [28] are known to lead to embryonic death in mice (after E18.5, E7.5 and E9.5, respectively) upon disruption. Pold absence caused lethality at an earlier embryonic stage than these aforementioned DNA polymerases. This is consistent with the notion that Pold has a fundamental role in cellular proliferation.
Genetic studies in yeast [4] and in vitro biochemical studies [8,9] have shown that Pold is indispensable for nuclear DNA replicative synthesis. However, in this presented study, we found no obvious abnormality in functionally Pold1-null mouse embryos before implantation. Since we could count more than 60 nuclei in a one day cultured Pold1 2/2 blastocyst (data not shown), at least 6 times of cell division were performed between fertilization and the embryonic death. To investigate whether these embryos could proceed with pre-implantation development without Pold1 protein, we performed immunostaining of Pold1 2/2 blastocysts with anti-Pold1 antibodies. Pold1 protein was not only found in Pold1 +/2 and Pold1 +/+ embryos but also in Pold1 2/2 embryos, which had no apparent cell cycle defect (Fig. S1). In addition, published microarray data indicated that the Pold1 mRNA presents in mouse unfertilized eggs and early embryos where zygotic gene expression have not started [29]. These results suggested that maternal Pold1 stockpiles could support for proceeding pre-implantation development of Pold1-null embryos. Similar peri-implantation lethalities have been found when disrupting other components required for nuclear DNA replication such as FEN1 [30], CDC45 [31], and Psf1 [32]. Maternal compensation is also thought to explain why in zebrafish, Pold1null embryos developed to a late embryonic state [15].
Pold1 2/2 embryos showed a cell proliferation defect and a DNA synthetic defect in in vitro blastocyst outgrowths. These observations show that no redundant activity for Pold exists in mammalian development and affirm the known Pold cellular function requirement in chromosomal DNA synthesis. Our Pold results also show frequent spontaneous apoptotic cells and few mitotic cells in 3 day cultured Pold1 2/2 blastocyst outgrowths (Fig. 3B), indicating the existence of a DNA replication checkpoint in eukaryotic cells to guarantee an accurate transmission of chromosomal DNA to subsequent generations [33]. Disrupting DNA replication is known to activate this checking system, resulting in either arresting the cell cycle before mitosis initiation in order to have the disruption removed or tolerated, or causing apoptosis on cells that harbor the disruption [34]. Our observations in 3 day cultured Pold1 2/2 blastocyst outgrowths is an expected cellular response to the defective DNA replication caused by Pold disruption.
Although heterozygous Pold1 +/2 mice showed a reduction in the amount of Pold1 protein (Fig. 1H), there were no apparent developmental abnormalities. In addition, Pold1 2/2 embryos developed normally until the blastocyst stage despite zygotic gene expression starting at the two cell stage. These observations showed that in vivo, the Pold1 gene expresses more than the necessary amount of gene products needed for development and indicate that the gross quantity of Pold1 protein does not regulate the mammalian developmental process.
In contrast to the functional disruption seen with the Pold1 gene, it has been seen that homozygous deficient (Pold1 exo/exo ) mice for Pold 39-59 exonuclease activity develop normally, although they are more susceptible to tumors [16,17]. We found similar phenotypes in Pold1 exo/2 mice. Similar types of tumors were observed in both Pold1 exo/exo and Pold1 exo/2 mice lacking exonuclease activity. The tumorigenesis in Pold1 exo/2 mice revealed more clearly that one Pold1 exo allele was sufficient for increasing cancer susceptibility without Pold1 + allele and Pold 39-59 exonuclease activity expressed a dominant effect on the suppression of tumorigenesis.
Most of our knowledge about DNA replication has accumulated from biochemical studies and genetic studies on lower eukaryotes. But these models do not necessarily correlate well with higher eukaryotic systems because eukaryotic systems have many more types of differentiated cells with more complicated interactions. For example, a study of DNA polymerase genes during the development of mouse testis suggested that Pole did not participate in meiotic replication, though it is thought that this polymerase cooperates with two others, Pola and Pold, in DNA replication based on other models [35]. Additionally, reliability of DNA replication, in which Pold is involved, affects cancer susceptibility in higher eukaryotes.
Our present findings revealed the physiological functions of Pold during mammalian development by using Pold1 gene targeted mice, indicating that the molecular function of Pold is conserved at least in mammalian early embryogenesis and that its 39-59 exonuclease activity is necessary for tumor suppression. Further analysis on the physiological function of replicative DNA polymerases will shed light on the mechanism of DNA replication in mammalian systems, adding critical insight to genomic instability.

Animals
Animals were maintained in a specific pathogen-free space under a 12-h light/dark regime with access to food and water ad libitum. Experimental procedures were in accordance with the Guide for the Care and Use of Laboratory Animals of the Science Council of Japan and were approved by the Animal Experiment Committee of Osaka University.

Construction of the gene targeting vector
The targeting vector was cloned into pBlueScriptII (Invitrogen) and consisted of four fragments. The bacterial artificial chromosome (BAC) clone RP23-406H21 comprised the murine genomic Pold1 coding region. For homologous recombination, a 39 fragment (6.9-kb) from the intron between exons 10 and 11 to exon 25 was extracted with EcoRI and SacI digestion from the  BAC clone. The 59 fragment (1.6-kb) from exon 5 to the intron between exons 10 and 11 contained a mutation which changes the D400 codon to an alanine codon. This mutated fragment was generated by repeated PCR using the oligonucleotide 59-CAGAACTTTGCCCTCCCATACCTC-39 and its complementary oligonucleotide (alanine codon is underlined). Floxed fragments, which contained the 800-bp simian virus 40 (SV40) splice/polyadenylylation signal [36] and a neomycin resistance cassette for positive selection, were inserted between the 59 and 39 fragments for homologous recombination and in the same orientation relative to the Pold1 gene. A DT-A fragment for negative selection was placed outside the 59 fragment for homologous recombination [37]. The targeting vector was verified by restriction digestion and nucleotide sequencing.

Establishment of Pold1 exo/+ ES cell lines
To obtain Pold1 exo/+ ES cells, circular pCre-Pac plasmids were electroporated into Pold1 +/2 ES cells [20]. The candidate ES cells were selected transiently with puromycin (1 mg/ml; Sigma). Each genotype of these Cre-mediated recombinants was confirmed by PCR and Southern blotting.

Generation of Sycp1-Cre Tg mice
The plasmid to create Sycp1-Cre Tg mice was constructed from two fragments. One fragment was the promoter sequence from the murine Sycp1 gene, which expressed in male germ cells, including the region from 2737 to +87 relative to the transcription start site [38]. The other was the Cre coding sequence containing the nuclear localized signal and polyadenylation signal. Both fragments were inserted into the pBluescript II (Stratagene) between the SpeI and HindIII sites. Then the plasmid was digested by SalI and NotI and the purified fragment was microinjected into fertilized eggs derived from C57BL/6 mice. We screened the offspring for an insertion of the transgene by genomic PCR using the Cre specific primers (59-CTGAGAGTGATGAGGTTC-39) and (59-CTAATCGCCATCTTCCAGCAG-39).

Blastocyst culture and genotyping
All embryos were generated by natural mating. The morning of the day on which the vaginal plug was detected was designated as day E0.5. Embryos were collected on E3.5 or E4.5 by flushing the uteri with M2 medium (Sigma). For culture, embryos were then cultured in complete ES cell medium containing leukemia inhibitory factor. In BrdU treated embryos, embryos were additionally cultured in the ES cell medium supplemented with 10 mM BrdU for 3 hours.

Immunocytochemistry and apoptotic cell detection
Embryos were washed in phosphate-buffered saline (PBS) containing 1.5% bovine serum albumin (BSA), fixed in 4% paraformaldehyde in PBS for 30 min at 4uC, and then permeabilized for 20 min at room temperature in PBS with 0.3% Triton X-100 and 1.5% BSA. For BrdU-treated embryos, DNA was denatured after permeabilization with 0.25 N HCl and 0.5% Triton X-100 for 20 min at room temperature and washed extensively in PBS with 1.5% BSA. Embryos were incubated with specific primary antibodies overnight at 4uC. The primary antibodies used in this study were mouse anti-BrdU (Dako Cytomation) and rabbit antiphosphohistone H3 (Ser-10) (Cell Signaling). The fluorescencelabeled secondary antibodies were purchased from Molecular Probes. TUNEL (terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling) assays to detect apoptotic cells were performed by using the In Situ Cell Death Detection Kit (Roche). Figure S1 Pold1 protein exists in Pold1 2/2 blastocysts. To check if Pold1 gene products are present in Pold1 deficient embryos, we immunostained blastocysts harvested from Pold1 +/2 intercross using two kinds of anti-Pold1 polyclonal antibodies, H300 and C20 (purchased from Santa Cruz) and counterstained with DAPI.