|Home | About | Journals | Submit | Contact Us | Français|
Mammalian genomes encode at least 15 distinct DNA polymerases, functioning as specialists in DNA replication, DNA repair, recombination, or bypass of DNA damage. Although the DNA pol zeta(ζ) catalytic subunit REV3L is important in defense against genotoxins, little is known of its biological function. This is because REV3L is essential during embryogenesis, unlike other translesion DNA polymerases. Outstanding questions include whether any adult cells are viable in the absence of Polζ, and whether Polζ status influences tumorigenesis. REV3L-deficient cells have properties that could influence the development of neoplasia in opposing ways: markedly reduced damage- induced point mutagenesis, and extensive chromosome instability. To answer these questions, Rev3L was conditionally deleted from tissues of adult mice using MMTV-Cre. Loss of REV3L was tolerated in epithelial tissues, but not in the hematopoietic lineage. Thymic lymphomas in Tp53−/− Rev3L conditional mice occurred with decreased latency and higher incidence. The lymphomas were populated predominantly by Rev3L null T-cells, demonstrating that loss of Rev3L can promote tumorigenesis. Remarkably, the tumors were frequently oligoclonal, consistent with accelerated genetic changes in the absence of Rev3L. Mammary tumors could also arise from Rev3L-deleted cells, in both Tp53+/+ and Tp53+/− backgrounds. Mammary tumors in Tp53+/− mice deleting Rev3L formed months earlier than mammary tumors in Tp53+/− control mice. Prominent preneoplastic changes in glandular tissue adjacent to these tumors occurred only in mice deleting Rev3L and were associated with increased tumor multiplicity. Polζ is the only specialized DNA polymerase yet identified that inhibits spontaneous tumor development.
Despite the cell’s many DNA repair mechanisms, genomic replication and cell division frequently take place with unrepaired DNA lesions still present. Bypass of such lesions can be performed by enzymes known as translesion synthesis DNA polymerases, which have activities on differing substrates (1, 2). Included in this group of enzymes in mammals are polζ, polη, polι, polκ, polθ, polν, and Rev1 (1, 2). The best characterized is pol η, which inserts bases across from UV radiation-induced cyclobutane pyrimidine dimers in DNA. A deficiency of the human enzyme causes the hereditary skin cancer- prone disorder xeroderma pigmentosum variant (XP-V) (1).
Polζ has been difficult to study because unlike other specialized DNA polymerases, its catalytic subunit REV3L is essential for development. This indicates that polζ has a non-redundant function, perhaps bypass of endogenous DNA damage in the rapidly growing embryo (3–6). The mammalian REV3L gene (7, 8) encodes a polypeptide of ~350 kDa, twice as large as the catalytic subunit of yeast polζ, Rev3. Reduced REV3L function in human and mouse fibroblasts decreases base change mutagenesis induced by UV light and other mutagens (9–11). REV3L therefore differs markedly from the other DNA polymerases, because interfering with its function reduces rather than increases point mutagenesis.
Polζ-defective cell lines have only been isolated in p53-compromised backgrounds. Rev3L null mouse embryonic fibroblasts exhibit spontaneous chromosomal instability, with an increase of ~10-fold in the frequency of chromosome translocations (12); other Rev3L null cells also have increased frequencies of chromosome breaks and translocations (13–16). This is consistent with a role for DNA polζ in bypass of adducts that would otherwise block DNA replication and lead to replication fork collapse and breakage. Alternatively, REV3L might function in the repair synthesis that occurs during some forms of DNA double strand break repair (16–19). We undertook a study to answer two basic questions. The first is if Rev3L loss can be tolerated in normal adult cells. A second is whether the loss of REV3L in adult cells affects the development of spontaneous neoplasia.
The conditional targeting construct (Fig. 1A) was based on that used to replace two exons containing the 3′ end of motif I and all of motif V within the REV3L polymerase domain active site (5). The plasmid backbone (pBluescript II KS+) and the Rev3L genomic homology arms were retained and loxP sites were added at the internal edges of the homology arms. At the 3′ end of the 5′ homology arm (BspE1 site), a neomycin resistance gene surrounded by loxP sites was added. This cassette was excised from plasmid pL2-neo (created by Hua Gu), contained a 1.1 kb linker regions and is expressed by the HSV tk promoter. A loxP site formed by annealing oligonucleotides was ligated to the 5′ end of the 3′ homology arm (SalI site). Finally the 4.6 kb of genomic sequence missing between the homology arm sequences and containing the essential polymerase exons was amplified from 129/Ola genomic DNA using Expand High Fidelity polymerase (Roche). This PCR product was cloned between the 5′ neo cassette flanked by loxP sites and the single 3′ loxP site. The final targeting construct contained 3 loxP sites and 4 exons (2 within the original 3′ homology arm and the 2 essential pol exons now flanked by loxP sites). The sequence of the exons and loxP sites was confirmed by DNA sequencing. It is not currently known whether any truncated REV3L protein is produced by this conditional construct.
The 3loxP targeting construct was linearized and electroporated into 129/Ola ES cells. Two ES cell clones were found to be correctly targeted (both 5′ and 3′ arms recombined) by nested PCR using decreasing temperature annealing steps. Southern blotting was performed to confirm this and to ensure that a second non-homologous integration had not occurred. The neo selection cassette flanked by loxP sites was removed by transient expression of a Cre recombinase plasmid within the ES cells. Individual colonies were then replica plated with one plate grown in the presence of G418 to identify those cells likely to have deleted the neo cassette. PCR screening was then employed to distinguish between colonies that had deleted the neo cassette only (to create the lox construct) versus those that also deleted the Rev3L exons because the outermost loxP sites had been utilized by Cre (creating a null allele). Southern blotting confirmed results and eliminated mixed ES cell colonies.
Two ES cell clones with the two essential DNA polymerase exons flanked by loxP sites were injected into C57BL6/J blastocysts. One clone generated pups with a transgenic germline. The Rev3L and Tp53 (20) strains were mixed B6;129. Mouse strains Tg(MMTV-Cre)1mam and Tg(MMTV-Cre)4mam, originally obtained from the Jackson laboratory on a FVB background, were bred to the Rev3L+/− strain (5). Mice were maintained as outbred by non-sibling matings. All experimental crosses were Rev3Llox/lox to Rev3L+/− MMTV-Cre, including those examining the effects of a Tp53 null background. Details of animal monitoring, necropsy and histology are in Supplementary Information.
A PCR assay was developed to determine the relative levels of conditional allele deletion in tissues and tumors of Rev3L+/lox Cre and Rev3L−/lox Cre mice. Genomic DNA was prepared from tissues and tumors using DNeasy columns (Qiagen). Keratinocyte culture and bone marrow lineage separation methods are in Supplementary Information. Triplex PCR used a forward primer common to both the lox and Δlox alleles and reverse primers unique to each allele. PCR reaction mixtures (25 μl) contained 200 μM each dNTP, 0.6 μM common primer, 0.3 μM of each allele-specific primer, 25–100 ng genomic DNA, and 0.13 μl Qiagen HotStar TaqPol (0.63 units). The annealing temperature (57 °C) was optimized to reflect defined input ratios of lox and Δlox genomic DNAs obtained from pure ES cell clones. After 30 cycles PCR, the 647 bp (lox) and 511 bp (Δlox) PCR products were separated on a 1.2% agarose gel.
Because of the embryonic lethality of Rev3L deletion, we developed a conditional knock-out mouse to determine the consequence of Polζ deletion for viability of different adult cell types and spontaneous tumorigenesis. Two essential exons in the DNA polymerase domain of the Rev3L gene were flanked by loxP sites so that they could be deleted by Cre recombinase (Fig. 1A). No differences in phenotype were apparent among Rev3L+/+, Rev3L+/−, Rev3L+/lox and Rev3L−/lox mice. To conditionally inactivate Rev3L, the lines Tg(MMTV-Cre)1mam and Tg(MMTV-Cre)4mam were used in crosses (23). In adults, both show highest levels of Cre expression in secretory tissues, the hematopoietic system (including B and T cells, megakaryocytes, and the erythroid lineage), and epithelia. Some MMTV-Cre expression may occur in embryos after 11.5 days of development (23). The life spans and weights of adult mice deleting the conditional lox allele to generate either Rev3L heterozygous (+ /Δlox) or Rev3L null (−/Δlox) cells were not statistically different from non-Cre controls (Fig. S1A).
A PCR assay was developed to quantify deletion in genomic DNA from Cre-expressing tissues (Fig. 1B). Significantly, deletion of Rev3L to null status was tolerated in epithelial tissues, including samples from the tail, ear and salivary gland, though probably with some loss of viability (Fig. 1B). Rev3L null cells were still present in these tissues in two year old Rev3L−/loxCre mice. We confirmed that Rev3L null keratinocytes are viable and can divide ex vivo (Fig. S2A).
The consequence of Rev3L deletion was quite different in hematopoietic tissues. In Rev3L−/loxCre mice, the deleted form of the conditional Rev3L allele was not detected or was present in only trace amounts in bone marrow, spleen, thymus and lymph nodes (Fig. 1C). The conditional Rev3L allele was consistently disrupted at a level of ~50% in the hematopoeitic tissues of all Rev3L+/loxCre control mice tested (Fig. 1C). Poor viability of Rev3L-deleted lymphoid cells was not a consequence of the introduction of double-strand breaks mediated by V(D)J recombination during the maturation of lymphocytes (Fig. S2B). This demonstrates an essential requirement for pol ζ in maintaining viability in adult hematopoietic lineages.
A lack of Rev3L could accelerate the formation of oncogenic rearrangements or inactivation of tumor suppressors. Lymphomas were the most common tumor in Tp53+/+ Rev3L−/loxCre mice and occurred with frequencies similar to the control genotypes. Most did not contain Rev3L null cells (Fig. S3A). Nevertheless, among the hematopoietic masses examined from 8 Rev3L−/loxCre mice, one T-lymphoblastic lymphoma/leukemia occurred very early (3.5 months) and was dominated by Rev3L null cells (Fig. 2A, S3B). Thus complete loss of Rev3L was an early event in this tumor and the final change(s) required for uncontrolled growth occurred in a cell lacking Rev3L function.
Mice were placed in a Tp53-null background, where the influence of Rev3L disruption on the frequency and latency of spotaneous thymic lymphoma could be measured. Cellular lethality upon Rev3L deletion is partially due to p53-imposed controls (12, 14, 15, 24). Deletion of Tp53 did not globally allow blood cells to proliferate without Rev3L function, as there was near or total absence of cells with complete disruption in the bone marrow, spleen and mesenteric lymph node of Rev3L−/loxTp53−/−Cre mice (Figs. S4A and B). However, examination of Rev3L deletion in the thymus of two Rev3L−/loxTp53−/−Cre4mam mice without obvious lymphoma revealed that absence of Tp53 did promote the ability of Rev3L null thymocytes to divide in vivo (Fig. S4A and B). In one case (Fig. S4B), Rev3L deleted cells were found only in a single thymic lobe, suggesting that absence of Tp53 is necessary but not sufficient for bypass of the genomic defect(s) induced by Rev3L loss.
Twenty-one of the 22 mice in the Rev3L−/loxTp53−/−Cre group succumbed to thymic lymphoma by 142 days (Table 1). The one that did not was euthanized at six weeks due to exencephaly known to occur in Tp53 null mice (25). The incidence of thymic lymphomas in the Rev3L−/loxTp53−/−Cre1mam group was thus 100%, revealing complete penetrance in mice capable of forming Rev3L null lymphocytes. This is significantly greater than the 70 – 77% frequency in control genotypes (p<0.03, Table 1), a frequency consistent with other studies of Tp53−/− mice (20, 26). Eleven tumors from Rev3L−/lox Tp53−/−Cre1mam mice were examined histologically and all were precursor T-lymphoblastic leukemia/lymphomas, based on CD3+TdT+ immunostaining and morphology. This is the predominant form of lymphoma arising in Tp53 null mice (27).
Mice from the Rev3L−/loxTp53−/−Cre group forming Rev3L null lymphomas have significantly decreased survival compared to the three control genotypes retaining at least one functional Rev3L allele (Fig. 3; p<0.0002, Table 1). Similar differences between the deleting and non-deleting genotypes were found by comparing only the survival of mice with lymphoma (Table 1). Genotyping multiple samples from 14 of the lymphomas from the Rev3L−/loxTp53−/−Cre1mam group showed that 10 were predominantly constituted by Rev3L null cells (representative results, Fig. 2B). One lymphoma had a mixed population that was mostly Rev3L null, but had one thymus sample also containing parental, non-deleted Rev3L cells (Fig. 2C). Three lymphomas contained only cells with the intact conditional allele. Thus, loss of Rev3L enhances development of thymic lymphoma in Tp53 null mice both by increasing incidence and by shortening the latency of tumors.
To test whether Rev3L− cells have an increased probability of neoplastic conversion, we asked whether thymic lymphomas arising from them are multiclonal. T-cell lymphomas from Tp53 null mice have clonal Tcrb rearrangements (28). Clonality in thymic lymphomas was assessed by determining the number, distribution and sizes of D-J rearrangements within the Tcrb (TCR-β) gene (22) in 10 Rev3L+ and 10 Rev3L− thymic lymphomas. To assess whether cell populations differed across the tumor, multiple genomic DNA samples from non-adjacent sites of each tumor were assayed simultaneously. D1-J1, D2-J2 and D1-J2 joined segments were amplified in separate PCR reactions (Fig. 4A) that paired Rev3L− tumors with Rev3L+ tumors (Fig. 4B, C, D). Non-tumor thymus DNA from a Tp53 null mouse served as a control to identify the sizes of normal rearrangements, all of which can be found in a normal thymus, including those involving a J2 pseudosegment (Fig. 4C, D, right).
From the ten Rev3L+ tumors, 48 total independent samples were analyzed; only one instance was found with different clonal proliferations at different locations within the tumor (Fig. 4B). Analysis of 39 samples from different sites of the ten Rev3L− tumors revealed that five Rev3L− tumors had different clonal populations in separate samples from the same tumor (Fig. 4B, C, D).
D-J recombination can occur on both alleles of the Tcrb locus (29). To classify a tumor sample as biclonal based on number of rearrangements, more than two J1 or J2 rearrangements must be present in an individual sample. No Rev3L+ tumor samples had so many products, whereas three Rev3L− tumors had more than two D2-J2 rearrangements in an individual sample. One of these was not among the five having different clonal populations at non-adjacent sites within the same tumor. The analyses together identify a significantly increased number of Rev3L null tumors that are at least biclonal (Fig. 4B).
The Rev3L− lymphoma group also had five D-J rearrangement products with sizes distinctly different than the standard rearrangements (Fig. 4B, C, D); only one such rearrangement occurred in the Rev3L+ group (Fig. 4B). The occurrence of aberrantly migrating D-J rearrangements suggests that processing of VDJ intermediates might be altered in Rev3L null T-cells. Rev3L may also be involved in the repair of strand breaks occurring during class switch recombination (16).
Mammary tumors were also observed in Rev3L-deleting females in the Tp53+/+ cohort (Table S2). Notably, 5 of the 6 mammary tumors in Rev3L−/loxCre females were composed largely of Rev3L null cells (Fig. S5A). The single mammary tumor (a solid carcinoma) from the Rev3L+/loxCre genotype had no Rev3L deletion (Fig. S5B).
To further examine the effect of Rev3L deletion on tumor incidence and latency, we generated cohorts of Rev3L-deleting and control genotypes in a Tp53-sensitized background. On the Tp53+/− background, significantly more mammary tumors occurred in mice deleting Rev3L than those retaining Rev3L function (Table S3, p=0.044). Eighteen mammary tumors were found in 11 Rev3L−/loxTp53+/−Cre1mam mice, with 4 of these in males. The numbers of females with mammary tumors from the Tp53+/− and Tp53+/+ cohorts were combined for each Rev3L genotype (although tumor numbers were smaller in the Tp53+/+ cohort). The number of Rev3L-deleting females that developed mammary tumors was significantly greater than the combined Rev3L+/loxCre control group, which had the next highest incidence (14 vs. 5 females, 41 total females in each group, p=0.035, Fisher’s exact test).
Four Rev3L−/loxTp53+/−Cre1mam mice but only a single control mouse developed multiple mammary tumors. As in the Tp53+/+ cohort, most mammary tumors (12/17 analyzed) in Rev3L-deleting mice were comprised largely of Rev3L null cells; samples from 2 other tumors contained 25–50% deletion. The increased number of mammary tumors populated by Rev3L null cells in Rev3L-deleting females strongly suggests that loss of REV3L promotes mammary tumorigenesis.
Mammary tumors in Rev3L deleting mice had a shorter latency than in control genotypes in the Tp53+/− background (median 295 vs. 448 days, p=0.02, log-rank). Loss of Rev3L thus reduced tumor latency in addition to increasing tumor incidence. Survival of Rev3L−/loxTp53+/−Cre1mam mice as a group was not significantly different from controls (Fig. S1B). Similar numbers of lymphomas developed in all 4 Tp53+/− genotypes (22–34%), comparable to previously reported incidences (26).
Mice on the Tp53+/− background, regardless of Rev3L status, developed a similar spectrum of mammary tumors, including carcinosarcomas, adenosquamous carcinomas, solid carcinomas, and sarcomas (Table S3). Half of the mammary tumors in Rev3L−/loxTp53+/−Cre1mam mice were solid carcinomas (compared to 1 of 7 mammary tumors in the 3 control genotypes); these tumors generally had the highest proportion of Rev3L null cells among the mammary tumors from Rev3L deleting mice.
Glandular tissue adjacent to the mammary tumors from Rev3L-deleting mice often had mammary intraepithelial neoplasia (MIN) and marked atypical hyperplasia (Fig. 5A). In these mice, MIN was observed adjacent to at least 7 tumors of different types. The 6 control genotype tumors with adjacent glandular tissue had no MIN and only the single control mouse with multiple mammary tumors had mild to moderate hyperplasia (Fig. 5A). In the Tp53+/+ cohort, MIN was found adjacent only to one Rev3L null adenosquamous carcinoma (Rev3L−/loxCre1mam female). Atypical hyperplasia and MIN would clearly predispose Rev3L-deleting mice to the development of multiple mammary tumors. Moreover, some mammary tumors in Rev3L-deleting mice may have been formed by the confluence of multiple neoplastic foci, as 3 of the 4 Rev3L−/lox Tp53+/−Cre1mam mice with multiple tumors (Fig. 5B) also had multiple MIN adjacent to the tumors. Thus loss of Rev3L in a Tp53+/− background causes an increase in the number of both preneoplastic and neoplastic lesions.
The conditional Rev3L knockout mouse permitted a test of the viability of tissues lacking polζ function. Epithelial cells from the tail, ear and salivary gland tolerated deletion of Rev3L, and null keratinocytes underwent at least 3–4 cell divisions in culture. In contrast, Rev3L null hematopoietic cells did not proliferate, consistent with the severe early hematopoiesis defect previously found in Rev3L null embryos (6). Hematopoietic tissues may be especially dependent on Rev3L for their survival, due to high basal levels of p53 (30), a higher load of reactive oxygen species (31), or a greater dependence on polζ-dependent DNA replication and repair. Removal of p53 did not enable viability of all blood cells, but promoted survival of Rev3L null thymocytes. This suggests that a Tp53-independent response is present, but that it can be bypassed or inactivated. Both Tp53-dependent and Tp53-independent pathways exist to induce cell cycle arrest, senescence, stasis, and apoptosis (32–34).
Remarkably, REV3L-negative tumors grew aggressively, although Rev3L is essential for embryonic development and viability of MEFs and blood cells. In Tp53−/− mice, deleting the conditional Rev3L allele reduced lifespan and increased thymic lymphoma incidence to 100%. Most lymphomas consisted primarily of Rev3L null cells, indicating that REV3L loss accelerates lymphoma development. Loss of Rev3L reduced mammary tumor latency in Tp53+/− mice, increased tumor multiplicity, and promoted the occurrence of mammary tumors in males. For both T-cell and mammary tumorigenesis, Rev3L loss apparently increases the number of cells at risk for neoplastic conversion. More Rev3L− tumors were oligoclonal, and mammary intraepithelial neoplasia was a distinct feature of adjacent glandular tissue in Rev3L null mammary tumors.
Genomic instability due to Rev3L loss has been previously documented in chicken, mouse and human cells, with increased formation of DNA breaks and translocations (12–16). This provides a likely explanation for enhanced tumorigenesis in Tp53-deficient mice deleting Rev3L. Increased genetic instability is known to accelerate tumorigenesis, although the generation of DNA breaks may require that Tp53-mediated controls be removed or attenuated.
The precise cellular function of REV3L in limiting genome instability remains to be determined. One possibility is that DNA polζ has a role in bypass of some endogenously formed adducts that would otherwise block DNA replication and lead to replication fork collapse and breakage. Alternatively, REV3L might function in the repair synthesis that occurs during DNA double strand break repair. The mid-gestation lethality of Rev3L deletion and severe proliferation defect of null embryonic fibroblasts somewhat resembles knockouts of genes involved in homologous recombination (HR), such as Rad51, Rad51d, Xrcc2, and Brca (3). Rev3L and HR null knockouts also share features of chromosome instability and Tp53 absence enables cells lacking these genome maintenance genes to overcome blocked cell division (35). Deletion of Brca1 or Brca2 in Tp53+/+ mice by MMTV-Cre results in mammary tumors with average latencies similar to that caused by Rev3L deletion (36, 37). Tp53 heterozygosity cooperates with the loss of all 3 of these genes to reduce the average latency of mammary tumor formation to between 8.5 months (Brca1) and 10.5 months (Rev3L). Deletion of Rev3L by MMTV-Cre in Tp53+/− mice induced multiple mammary tumors, as it did for the (more penetrant) Brca1 (36). Deletion of the Brca alleles in Tp53−/− mice using the T-cell lineage specific Lck-Cre (38, 39) has effects on incidence and latency of thymic lymphoma similar to the acceleration of T-cell lymphoma caused by Rev3L deletion.
The present results with Rev3L identify a specialized DNA polymerase that uniquely inhibits spontaneous tumor development in mammals The finding that epithelial cells can survive in the absence of pol zeta opens the way to further investigations of the specific roles of this enzyme in epithelial neoplasia, including UV-induced skin cancer.
Expert assistance with genetically modified mice was provided by Ian Rosewell and Steve Wilson (CRUK transgenic services) and J. R. Chaillet (U. Pittsburgh). Advice on statistical analysis was provided by Hong Wang (U. Pittsburgh) and Howard Thames and Kevin Lin (M. D. Anderson). We thank Gregory Gan for discussion and our colleagues for comments on the manuscript. This work was supported by NIH grants CA132840 and CA098675 (to RDW), NIH award K08 AR053566 to LJR, the Hillman Foundation (JPW), grant P30-ES007784 from the National Institute of Environmental Health Sciences, and NIH Cancer Center Support Grants P30-CA016672 (University of Texas M. D. Anderson Cancer Center) and P30-CA47904 (University of Pittsburgh Cancer Institute).