Translesion Synthesis Polymerases in the Prevention and Promotion of Carcinogenesis

A critical step in the transformation of cells to the malignant state of cancer is the induction of mutations in the DNA of cells damaged by genotoxic agents. Translesion DNA synthesis (TLS) is the process by which cells copy DNA containing unrepaired damage that blocks progression of the replication fork. The DNA polymerases that catalyze TLS in mammals have been the topic of intense investigation over the last decade. DNA polymerase η (Pol η) is best understood and is active in error-free bypass of UV-induced DNA damage. The other TLS polymerases (Pol ι, Pol κ, REV1, and Pol ζ) have been studied extensively in vitro, but their in vivo role is only now being investigated using knockout mouse models of carcinogenesis. This paper will focus on the studies of mice and humans with altered expression of TLS polymerases and the effects on cancer induced by environmental agents.


Introduction
Tumorigenesis is a multistep process beginning with the transformation of a single cell by the accumulation of at least six distinct characteristics. These include infinite lifespan, resistance to antigrowth signals, resistance to apoptosis, autocrine production of growth signals, sustained angiogenesis, and tissue invasion [1]. Most environmental carcinogens induce transformation by causing mutations in the DNA that alter the activity of protooncogenes or tumor suppressors. These mutations are formed when residual, unrepaired DNA damage stalls progression of the replication fork during S phase. Stalled replication forks are most frequently resolved using error-free mechanisms that include homologous recombination or use of the homologous nascent strand as a template. Nevertheless, replication may proceed using the damaged strand as a template in an errorprone process known as translesion DNA synthesis (TLS). TLS is defined as the incorporation of a nucleotide across from DNA damage followed by extension of the potentially mispaired primer-template, and can be error-free or errorprone. Cellular commitment to error-free, recombinatorial damage avoidance or error-prone TLS is modulated by the molecular switch PCNA (Figure 1). Cells presumably risk mutations caused by TLS to relieve replication fork blockage at DNA adducts and to avoid the potential formation of extremely cytotoxic double strand breaks (DSB). Although it accounts for less than 10% of all bypass synthesis events in yeast [2], the frequency of potentially mutagenic TLS may be as high as 50% in higher eukaryotes [3][4][5]. The propensity and mutagenic potential of TLS explain why it is etiologic in most environmentally-induced cancers and has been the focus of numerous investigations over the past decade.
TLS is performed by a relatively new category of accessory DNA polymerases. Polymerase η (Pol η), Pol ι, Pol κ, and REV1 in the Y-family [6] and Pol ζ in the B-family [7,8] are responsible for most TLS in mammalian cells. These proteins have active sites that are larger and more open than those of the high-fidelity replicative DNA polymerases (Pol α, δ, and ε), allowing accommodation of and synthesis past 2 Journal of Nucleic Acids DNA templates with large, helix-distorting lesions [9]. This unique ability to synthesize DNA opposite bulky adducts helps cells avoid double strand breaks associated with replication fork stalling, but can also lead to mutagenesis by incorrect base addition. It is important to note that polymerases in the Y-family are expressed in all three kingdoms of life, indicating a critical and evolutionarily conserved role for these proteins [6]. The obviously conflicting roles of these enzymes in both preventing and promoting genetic instability are reflected in the tight cellular control of the TLS pathway ( Figure 1). Although extensive in vitro studies have given us a better understanding of their role in the cell, much less is known about the function of TLS polymerases in living animals. Limited epidemiological studies have been conducted to associate single nucleotide polymorphisms (SNPs) with cancer risk in humans. Knockout mice have been generated for each gene, and carcinogenesis studies are published or underway. Importantly, studies in mice and humans have shown that TLS polymerases, particularly Pol η, are involved in immunoglobulin gene hypermutation. Readers are directed to reviews by Reynaud et al. and Diaz et al. for an exploration of this function of TLS polymerases [10,11]. This review will focus on the rapidly progressing connection of TLS and cancer research in knockout mice and human populations.

REV1
REV1 was discovered in budding yeast by the Lawrence group in 1989 as a component of the Pol ζ complex [17]. The catalytic activity of REV1 is limited to insertion of dCMP across from a template dG [18]. The human homolog was cloned in 1999 and has the same template-dependent dCMP transferase catalytic activity on an undamaged template or an abasic site [19]. One locus used in eukaryotic cells to measure mutation frequency is HPRT, a gene involved in the purine salvage pathway. In this forward mutation assay, cells with loss-of-function mutations in HPRT are resistant to the drug 6-thioguanine (TG). REV1 is required for carcinogeninduced HPRT mutagenesis in human cells [20][21][22], but the catalytic activity appears to be dispensable for the induction of UV-induced mutations [23,24], indicating that this protein probably plays a structural rather than catalytic role in UV mutagenesis.
Cells from mice with a targeted deletion of the BRCA1 Cterminal (BRCT) homology domain of Rev1 (Rev1 B/B ) have a reduced UV-induced mutation frequency at the Hprt locus [25]. However, the animals have a paradoxically decreased latency of squamous cell carcinoma (SCC) formation and only marginally reduced p53 mutagenesis in the skin after UV exposure [26]. Despite Rev1 B/B cells showing a moderate increase in chromatid breaks and exchanges after UV in vitro [25], comparative genomic hybriziation of UV-induced SCC and normal skin DNA reveals no increase in the frequency of gross genomic alterations in Rev1 B/B SCC [26].  Figure 1: Regulation of DNA lesion bypass in Saccharomyces cerevisiae and humans. Bulky DNA lesions can cause blockage of replicative polymerases and replication fork stalling. The ubiquitin conjugase/ubiquitin ligase pair Rad6/Rad18 is recruited to stalled replication forks where the proteins catalyze monoubiquitylation of PCNA at lysine 164. TLS proteins such as REV1 and Pol η have increased affinity for monoubiquitylated PCNA, which facilitates their recruitment and the completion of TLS. In yeast, Rad5 and the MMS2-Ubc13 complex (UBE2V2-UBE2N in humans) can catalyze polyubiquitylation of PCNA via lysine 63 of ubiquitin, which blocks TLS and activates error-free damage avoidance. Damage avoidance includes template switching, during which the nascent DNA strand from the sister duplex is used as an undamaged homologous template to replicate past the lesion. Humans express two Rad5 homologs, SHPRH and HLTF, and both catalyze K-63-linked polyubiquitylation of PCNA in human cells [12][13][14][15]. In yeast, Ubc9-Siz1 can attach the small ubiquitin-like modifier (SUMO) to lysine 164 of PCNA in a reaction that competes with Rad6/Rad18-mediated monoubiquitylation. PCNA SUMOylation at K-164 attracts the helicase Srs2 and prevents error-prone RAD52-dependent recombination. Reproduced with permission from Watson et al. [16]. of these Rev1-mutant mice induces enhanced Atr signaling, senescence, and apoptosis in the skin. However, long-term low-dose UV exposure causes a mitogenic response, as evidenced by epidermal hyperplasia, decreased apoptosis, and increased proliferation of CPD-containing keratinocytes [26]. Based on literature reports of the etiological role of IL-6 in carcinogenesis and elevated IL-6 levels in the skin after a single subtoxic UV dose, the authors conclude that errorprone TLS of UV-induced DNA damage is responsible for suppressing the proinflammatory, tumor-promoting effects of UV in the skin. However, more direct immunological studies are needed to confirm that Rev1 suppresses UVinduced inflammation and tumor suppression.
REV1 has also been implicated in TLS across other types of DNA lesions. Benzo[a]pyrenediolepoxide (BPDE) is the primary carcinogenic metabolite of B[a]P and causes point mutations in a REV1-dependent manner [22,27]. BPDE-induced Hprt mutations are dramatically decreased in primary mouse fibroblasts after ribozyme-mediated Rev1knockdown. When a plasmid expressing this ribozyme is delivered to the lungs of A/J mice by aerosol nebulization, Rev1 mRNA is reduced by ∼ %50 in the bronchial epithelium. This targeted gene therapy causes a ∼ %40 reduction in the lung tumor multiplicity after B[a]P treatment. In addition, only 73% of ribozyme-treated mice develop lung adenomas after B[a]P, compared with 100% penetrance in control animals [28]. This report highlights the potential for interrupting translesion synthesis as a chemoprevention strategy.
Although no human disorder involving REV1 deficiency is known, there are 16 SNPs in humans that result in nonsynonymous amino acid changes. The F257S SNP, which lies outside of all known functional domains of the protein, has been associated with an increased risk of squamous cell carcinoma of the lung in patients who have ever smoked cigarettes [29]. However, this association remains controversial [30]. The same F257S SNP was associated with decreased risk of cervical cancer, and N373S within the catalytic domain was associated with increased risk of cervical cancer. Both effects were specific for squamous cell carcinoma and not relevant for adenocarcinoma of the cervix [31]. Although the functional consequences of these polymorphisms are unknown, these studies support a role for REV1 in the formation of multiple internal cancers.

Pol η
The study of translesion synthesis in mammals began in 1999 with the discovery of the molecular defect that results in Xeroderma Pigmentosum (XP) variant syndrome. All XP patients have dramatically increased susceptibility to UV-induced skin cancer [32]. Patients in complementation groups A through G are deficient in nucleotide excision repair (NER), the major pathway for removal of helixdistorting lesions, including those induced by UV. However, the XP variant subset of patients has normal NER activity [33,34], yet displays the skin cancer-prone phenotype of NER-deficient patients. The XP variant mystery persisted for nearly three decades. Intensive investigations indicate that after UV-irradiation cells from these patients have difficulty exiting S-phase that is exacerbated by caffeine [35,36]. Further, these cells are extremely hypermutable after UV [37]. In 1999, two groups independently discovered that XP variant patients carry autosomal recessive mutations in POLH, the human gene coding for Pol η, and that the enzyme can catalyze error-free DNA synthesis across from a template TT cyclobutane pyrimidine dimer (CPD) [38][39][40]. The dramatic increase in skin cancer risk of XP variant patients could now be explained by the absence of a critical translesion DNA polymerase. UV principally induces photoaddition products between intrastrand adjacent pyrimidines, the most frequent of which are TT CPD. These lesions block progression of the replication fork. Data indicate that helicase activity may continue in spite of the blocked replication complex, resulting in single-stranded DNA that is rapidly coated with replication protein A (RPA). This appears to attract the ubiquitin ligase RAD18, which has binding sites for the ubiquitin conjugase RAD6, Pol η, and RPA. One target of ubiquitylation is PCNA. Since Pol η has a ubiquitin binding domain, Pol η is now thought to be preferentially attracted to the stalled fork because it is chaperoned directly by RAD18 and binds to the ubiquitylated PCNA ( Figure 1) [41]. Data indicate that Pol η then incorporates AA across from TT CPD in the template. In the absence of Pol η, another translesion polymerase, which is potentially error-prone when bypassing these common UV-induced lesions, accesses the damaged template (reviewed in [42,43]). Generation of Pol ηknockout mice shows that the highly homologous mouse Pol η protein functions similarly in UV-induced mutagenesis and carcinogenesis. Pol η-deficient mice develop squamous cell carcinoma with 100% penetrance at a UV fluence that does not cause any tumors in wild-type littermates. In addition, approximately one-third of heterozygous mice develop cancer after UV exposure [44]. This raises the possibility that humans carrying heterozygous mutations in the POLH gene may have an increased risk of developing skin cancer. However, this speculation has not been clinically investigated.
There is evidence that XP variant patients develop internal cancers faster than Pol η-proficient individuals [45,46], raising the possibility that Pol η-deficiency is involved in the formation of multiple human cancers caused by DNA damaging agents other than UV. Six SNPs in POLH have been found to date that result in nonsynonymous amino acid substitutions, but their functional significance is unknown. There is a single study evaluating the effects on cancer risk of POLH polymorphisms. Flanagan and colleagues found no significant changes in coding-region SNPs of POLH among 40 basal cell carcinoma and squamous cell carcinoma patients in a fair-skinned Irish population [47]. It is clear that larger epidemiological studies of POLH status are needed to evaluate the effects of POLH polymorphisms.

Pol ι
DNA polymerase ι (Pol ι) was discovered in 1999 as a novel homolog of Pol η in mammals and is encoded by the human POLI gene [48]. In vitro studies with purified enzyme indicate error-prone TLS function on almost all substrates examined, perhaps due to the still controversial ability of Pol ι to incorporate incoming nucleotides using Hoogsteen base pairing [49,50]. Exhaustive characterization of the error-prone replication properties of Pol ι has lent credibility to the hypothesis that Poli is a candidate gene for the Pulmonary adenoma resistance 2 (Par2) locus in mice [51][52][53]. The Par2 locus was identified in 1996 by chromosomal linkage mapping between BALB/cJ and A/J mouse strains and plays a major role in the relative resistance of BALB mice versus the A/J strain to developing urethaneinduced lung adenomas [54]. Wang et al. identified ten amino acid-substitution polymorphisms between A/J and BALB mice that produce changes in substrate recognition of Pol ι; while the enzyme from both strains is functional, the isoform expressed in BALB mice may be more accurate on certain undamaged templates [52]. These studies suggest that Pol ι acts to suppress urethane-induced lung adenomas. It has been hypothesized that this activity is due to the augmentation of base excision repair (BER) by Pol ι, because the enzyme has 5 deoxyribose phosphate (dRP) lyase activity and can partially reconstitute the BERdeficiency of Pol β-null cells in vitro [55]. It is possible that after urethane-induced DNA damage, which produces 1, N 6 -ethenoadenine adducts [56] that are primarily repaired by BER [57], Pol ι acts in the gap-filling step of lesion repair. If the isoform of Pol ι expressed in A/J mice is more likely to add the incorrect G opposite a template T in the gap-filling step of BER, as was found in vitro [52], this could explain the increased incidence of lung adenomas in A/J mice. In support of this hypothesis, nearly all urethane-induced adenomas in mice have a CAA → CGA transition in codon 61 of Kras2 [51]. In addition, 129derived mouse strains that carry a SNP in codon 27 of Poli resulting in a severely truncated protein [58] display extreme sensitivity to urethane-induced lung adenomas [53]. In the absence of Pol ι, it has been hypothesized that another DNA polymerase, such as Pol β, inserts the incorrect base during gap filling in the repair of urethane-induced DNA damage. However, normal mouse Pol ι displays extremely error-prone properties during synthesis opposite all four undamaged template bases in vitro [58], making it unlikely to prevent mutations during BER in mice that are Pol ιcompetent. Further studies must be completed to determine the tumor suppression mechanism of Pol ι in mouse lung carcinogenesis.
A growing body of evidence suggests that Pol ι is involved in error-prone TLS of UV-induced DNA damage in vivo. The heightened UV mutagenesis of Pol η-null (XP variant) human cells has been attributed to TLS by Pol ι [59]. Loss of the functional Poli gene in dermal cells results in a dramatically reduced UV-induced mutation frequency at the Hprt locus in both wild-type and Pol η-deficient mice ( Figure 2). Remarkably, however, the decreased UVinduced mutagenesis observed due to loss of the errorprone Pol ι from Pol η-deficient mice is associated with increased cancer risk after UV exposure ( Figure 3) [60]. This result was confirmed and extended by Ohkumo and colleagues who showed that Poli −/− mice are more likely to develop aggressive mesenchymal tumors after UV than Poli-proficient siblings [61]. These apparently contradictory findings speak to the fact that cancer etiology is more complex than the point mutations scored by the Hprt assay, and that one cannot use cell biology alone to accurately predict cancer risk in a TLS model. Indeed, they suggest a tumor suppressor role for Pol ι that could be separate from its role as a TLS polymerase prone to induce single base-substitution mutations. It is also possible that Pol ι is error-free when bypassing a minor UV adduct, or that it is involved in error-free BER of the minimal oxidative damage induced by UVB used in these studies [62], but more detailed experiments must be performed to evaluate these possibilities.
The role of Pol ι in the induction of cancer induced by other carcinogens has not been systematically studied to date. It is interesting to note that Newcomb et al. found that Pol ι-deficient 129 mice are resistant to γ-irradiationinduced thymic lymphoma but sensitive to methylating agent-induced thymic lymphoma [63]. γ-Irradiation induces DNA strand breaks and oxidative damage, and Pol ι is known to protect cells from oxidative stress [64]. It is therefore possible that increased cell death after γ-irradiation protects Poli −/− 129 mice from lymphomagenesis. However, Pol ι does not affect the sensitivity of Pol β-null cells to methylating agents [65], so the sensitivity of 129 mice to thymic lymphoma induced in this way is still unexplained. There is no known human disorder involving deficiency for Pol ι. However, Pol ι is overexpressed in some lung cancer cell lines [52] as well as in primary human gliomas [66]. The T706A SNP was found to increase the risk of adenocarcinoma and squamous cell carcinoma of the lung in persons < 61 years of age [29]. However, this association was not confirmed by another independent study [67] and failed to show significance in a meta-analysis [68]. Another SNP in human POLI, F532S, is associated with prostate cancer patients whose tumors display TMPRSS2-ERG fusion with a highly significant odds ratio of 4.6 [69]. The protooncogenic transcription factor ERG was identified as the most frequently overexpressed gene in human prostate cancers [70], and fusion with the androgen-responsive serine protease TMPRSS2 by chromosomal rearrangement was found in >90% of ERG-overexpressing cases [71]. Threonine 706 and serine 532, the two residues altered by these SNPs in Pol ι, are located in the noncanonical ubiquitin-binding motifs UBM2 and UBM1, respectively [72]. These two polymorphisms could therefore affect binding of Pol ι to ubiquitylated PCNA, which is required for its recruitment to stalled replication forks following DNA damage. In the case of prostate cancer, the F532S variant of Pol ι may promote chromosomal instability by causing replication fork stalling and double-strand break (DSB) formation. DSB formed in this way could promote cellular transformation by causing chromosomal rearrangements that place the protooncogene ERG under control of the androgen-responsive promoter elements of TMPRSS2 and lead to ERG-overexpression as is found in many prostate cancers [71]. Evidence supports the suppression of skin and lung cancers by Pol ι in humans and mice, and new studies suggest that other cancers could be affected by this protein, making it a promising candidate for future investigation.

Pol κ
The fourth member of the Y-family is DNA polymerase κ. Pol κ performs faithful TLS of BPDE-induced DNA damage in vitro by inserting dC opposite a template BPDEadducted G [73][74][75]. Pol κ is required for recovery from a novel BPDE-induced intra-S phase checkpoint, and the protein relocalizes to stalled replication forks after BPDEinduced DNA damage [76,77]. Polk −/− mouse embryonic fibroblasts (MEFs) show persistent S-phase arrest after BPDE exposure, which results in increased DSB formation at stalled replication forks and increased toxicity in cells without functional Pol κ [77]. Avkin and colleagues measured TLS efficiency and fidelity in Polk −/− MEFs using a shuttle vector technique. TLS efficiency on a plasmid containing a sitespecific BPDE-N 2 -dG adduct is reduced nearly threefold in Polk −/− MEFs, and mutagenic TLS is increased from 29% to 50% in knockout cells, supporting a role for Pol κ in the efficient and error-free bypass of BPDE DNA damage [78]. siRNA-mediated POLK-knockdown also reduces the efficiency of TLS past BPDE-N 2 -dG in human U2OS cells [79]. This body of evidence suggests that Pol κ could have an important role in cancers caused by bulky chemical carcinogens like BPDE. Pol κ has also recently been linked to nucleotide excision repair. Polk −/− MEFs have reduced levels of NER of UV damage, including reduced repair synthesis and removal of 6-4 photoproducts after UV. Both of these phenotypes are largely corrected by expressing wild-type Pol κ, but not a catalytically inactive mutant [80]. Pol κ carries out NER repair synthesis and is recruited to sites of NER through its interaction with XRCC1 and ubiquitylated PCNA [81]. These remarkable studies highlight the ability of TLS polymerases to function in multiple cellular pathways and the likelihood that Polk plays an important role in preventing DNA damage-induced carcinogenesis. While Polk-knockout mice have been generated [82,83] and show increased spontaneous mutagenesis in kidney, liver, and lung [84], no cancer studies have yet been reported using these models.
Pol κ is overexpressed in ∼70% of nonsmall cell lung cancers (NSCLC) examined [85], and this overexpression correlates with mutation status of TP53 [86] which is itself an indicator of poor prognosis [87]. In addition, POLK promoter activity is increased in TP53 −/− cells, and p53 protein suppresses POLK promoter activity in vitro. These reports suggest that Pol κ overexpression in NSCLC could be secondary to loss of functional p53, but the correlation between these two events must be investigated to rule out an etiological role for Pol κ in lung cancer. Much stronger epidemiological evidence shows that Pol κ is overexpressed in gliomas. Multivariate analysis indicates that Pol κ overexpression is an independent prognostic factor for the assessment of glioma patient outcomes (Figure 4) [66]. Although the potential role of Pol κ in the etiology of brain tumors is unclear, Pol κ is clearly a candidate for investigation of cancer risk and chemoprevention of multiple tumor types.

Pol ζ
The human homolog of yeast DNA Polymerase ζ is required for mutagenesis by UV, BPDE, and other carcinogens [7,88].
Pol ζ belongs to the B-family of DNA polymerases and contains a large catalytic subunit encoded by the REV3 gene in humans [7] along with the much smaller regulatory protein REV7 [89]. Early investigations in Saccharomyces cerevisiae showed that rev3 mutant strains have reduced rates of spontaneous mutation [90], indicating that Pol ζ is involved in the mutagenic processing of spontaneous and UV-induced mutations. Studies in mammalian cells indicate that Pol ζ has a role in both repair of double strand breaks and base substitution mutagenesis, the latter likely involving extension of mispaired primer termini after initial TLS by another polymerase [91,92]. Pol ζ is the only TLS polymerase required for development, and complete Rev3-knockout results in mitotic catastrophe and lethality at mouse embryonic day 10.5 [93][94][95]. However, conditional Rev3-knockout mice have been generated and are viable and fertile. While Rev3-deficiency alone is insufficient to promote cancer formation, conditional Rev3 knockout accelerates the spontaneous formation of lymphoma in Tr p53 −/− mice. In humans, REV3 gene expression is reduced by twofold in 40 of 74 (54%) colon carcinomas compared to matched normal tissue [96]. However, normal expression is found in much smaller sample sets of gastric, colon, lung, and renal cancers [97], and the gene is not mutated in primary tumors or cell lines from breast and colon cancers [89]. The expression levels of REV3 in human cancers, particularly colon carcinoma, must be revisited using larger sample sizes to draw firm conclusions about the correlation of gene expression and cancer progression. No studies are published investigating the 25 nonsynonymous SNPS in human REV3, but the possibility exists that functional changes in human Pol ζ could alter the risk of cancer formation.
Journal of Nucleic Acids 7

Conclusions
The importance of translesion DNA synthesis in preventing human cancer is well understood from the example of XP variant, in which patients lacking the Y-family DNA polymerase η are prone to develop UV-induced skin cancers due to an extremely hypermutable phenotype. However, we understand very little about how the other polymerases involved in TLS affect human health and cancer risk.
Recently developed mouse models have so far provided conflicting results; ribozyme-mediated knockdown of total Rev1 and removal of the BRCT domain both result in reduced mutagenesis by BPDE or UV, respectively. As expected, when Rev1 mRNA is knocked down using the same ribozyme delivered to the lungs of mice, multiplicity of B[a]P-induced lung adenomas decreases [28]. In contrast, Rev1 BRCT-null mice develop UV-induced squamous cell carcinomas faster than wild-type controls [26]. In a similarly paradoxical finding, mice lacking both Pol η and Pol ι have decreased UV-induced mutations in their dermal fibroblasts and accelerated development of squamous cell carcinoma after UV treatment compared to Pol ι-proficient animals [60]. While it is understood that the mutations induced by these polymerases are etiological in many environmentallyinduced cancers, it is clear from these studies that simply blocking TLS is not sufficient to reduce cancer risk, and in fact may cause an acceleration of carcinogenesis. More detailed studies are needed using existing mouse models to determine the effects of TLS polymerase activity on cancer development after diverse carcinogen exposures. In addition, molecular epidemiological studies must be conducted to evaluate the functional consequences of the many nonsynonymous SNPs in TLS polymerase genes, some of which have already been associated with cancer risk or protection. After a decade of intense research, there are still critical gaps in our understanding of the role of TLS in human health and cancer risk.