Adaptive use of error-prone DNA polymerases provides flexibility in genome replication during tumorigenesis

Abstract

Human cells possess many different polymerase enzymes, which collaborate in conducting DNA replication and genome maintenance to ensure faithful duplication of genetic material. Each polymerase performs a specialized role, together providing a balance of accuracy and flexibility to the replication process. Perturbed replication increases the requirement for flexibility to ensure duplication of the entire genome. Flexibility is provided via the use of error-prone polymerases, which maintain the progression of challenged DNA replication at the expense of mutagenesis, an enabling characteristic of cancer. This review describes our recent understanding of mechanisms that alter the usage of polymerases during tumorigenesis and examines the implications of this for cell survival and tumor progression. Although expression levels of polymerases are often misregulated in cancers, this does not necessarily alter polymerase usage since an additional regulatory step may govern the use of these enzymes. We therefore also examine how the regulatory mechanisms of DNA polymerases, such as Rad18-mediated PCNA ubiquitylation, may impact the functionalization of error-prone polymerases to tolerate oncogene-induced replication stress. Crucially, it is becoming increasingly evident that cancer cells utilize error-prone polymerases to sustain ongoing replication in response to oncogenic mutations which inactivate key DNA replication and repair pathways, such as BRCA deficiency. This accelerates mutagenesis and confers chemoresistance, but also presents a dependency that can potentially be exploited by therapeutics.

Abbreviations
• BRCA
• breast cancer gene
• CFS
• common fragile site
• DDT
• DNA damage tolerance
• DSB
• double-strand break
• EGFR
• epidermal growth factor receptor
• HR
• homologous recombination
• ICL
• interstrand crosslink
• NHEJ
• non-homologous end joining
• PCNA
• proliferating cell nuclear antigen
• Pol
• polymerase
• SNV
• single nucleotide variant
• TCGA
• The Cancer Genome Atlas
• TLS
• translesion synthesis
• TMEJ
• theta-mediated end joining

INTRODUCTION

DNA replication is the essential process for passing genetic information to the next generation and its accuracy and efficiency are critical factors in maintaining the integrity of the DNA sequence. In human bodies, the loss of replication fidelity leads to somatic mutations and, although most of them are harmless, their steady accumulation can eventually result in phenotypic consequences. Acquired mutations in a particular subset of genes represent driver mutations, which confer a selective growth advantage over cells in surrounding tissues and promote tumorigenesis. The mutation burden observed in tumor tissues varies between cancer types, with the number of acquired mutations ranging from 1000 to 20,000 in most cancers.¹ While it was commonly believed that the number of mutations primarily depends on the specific DNA repair pathway altered in a given cancer, a recent statistical analysis predicted that two-thirds of the mutations in human cancers result from replication errors,² highlighting the potential impact of DNA replication dynamics on mutator phenotypes in cancer cells.

In humans, DNA synthesis is conducted by the coordinated actions of at least 17 DNA polymerases, each with a distinct fidelity and efficiency to their activities. The accuracy of DNA synthesis by DNA polymerases is a key factor in determining the rate of mutagenesis and how each DNA polymerase is deployed during genome replication is therefore a primary factor in determining the overall stability of genome duplication.³ Eukaryotes encode polymerases from six families (A, B, X, Y, RT, and PrimPol; Table 1), which are generally categorized by their sequence homology or structural similarity to representative polymerases, for example the A-family is typified by DNA Pol I in Escherichia coli, and the B-family by Pol II. While, to some extent, the family of a DNA polymerase reflects its functions, polymerases within the same family can have diverse roles in DNA synthesis. It is critical that cells possess a functionally diverse ‘toolbox’ of polymerases since DNA is a labile and dynamic substrate. Non-canonical DNA conformations, including damaged bases, structures, and protein-bound DNA, have the potential to block the progression of polymerases or induce breaks in the DNA chain. These scenarios generate myriad potential substrate configurations that must be resolved or repaired by processing pathways involving the different polymerases. Thus, to avoid detrimental chromosome aberration and maintain cell fitness in any circumstance, cells must be able to balance the accuracy and flexibility of DNA synthesis.

TABLE 1 DNA polymerases of eukaryotes.

Family	Polymerase	Subunits	Major reported functions
A	Pol γ	POLG1, POLG2	Mitochondiral DNA replication
	Pol ν	POLN	interstrand crosslink repair/meiotoc homologous recombination
	Pol θ	POLQ	Translesion synthesis/theta-mediated end joining
B	Pol α	POLA1, POLA2	Bulk genome synthesis (lagging strand initiation)
	Pol δ	POLD1, POLD2, POLD3, POLD4	Bulk genome synthesis (lagging strand)
	Pol ε	POLE1, POLE2, POLE3, POLE4	Bulk genome synthesis (leading strand)
	Pol ζ	REV3L, REV7, POLD3, POLD4	Translesion synthesis (extension)
X	Pol λ	POLL	Non-homologous end joining
	Pol μ	POLM	Non-homologous end joining
	Pol β	POLB	Base excision repair
	Terminal deoxynucleotidyl transferase	DNTT	Non-homologous end joining
Y	Pol η	POLH	Translesion synthesis
	Pol ι	POLI	Translesion synthesis
	Pol κ	Pol κ-encoding gene	Translesion synthesis
	Rev1	REV1	Translesion synthesis (scaffold)
Prim-Pol	PrimPol	PRIMPOL	Repriming
Reverse transcriptase	Telomerase	TERT	Synthesis of telomeric repeats

Many recent studies have uncovered a high degree of plasticity in the usage of polymerases, and their functions in cells are far more diverse than previously anticipated. This becomes particularly relevant in cancer cells, where oncogenic activations change many aspects of cell physiology, including proliferation, metabolism, regulation of cell cycle, transcription, and epigenetics, which can each influence DNA replication processes, either directly or indirectly. Consequently, alterations to the functions of many polymerases are necessary, for example, to complete DNA synthesis in under-replicated regions or compensate for reduced activity of particular DNA repair pathways. In this review, we first summarize the canonical functions of DNA polymerases. We then examine how polymerase expression is misregulated in cancers, with a particular focus on several error-polymerases that have reported links with DNA replication in cancer. We then assess how the regulatory mechanisms of polymerases respond to oncogenic activation to influence polymerase usage and the overall fidelity of synthesis in cancer cells. Finally, we discuss the contribution of non-replicative polymerases to cancer adaptation following the loss or mutation of key replicative pathways and how this can provide opportunities for therapeutic intervention.

CANONICAL ROLES OF DNA POLYMERASES

The roles of many polymerases during canonical DNA replication are well established (Figure 1 and Table 1). Members of the B-family (Pol α, Pol δ, and Pol ε) are termed replicative polymerases, as they are the main contributors to synthesis in ongoing replication forks. Multiple studies in budding yeast and fission yeast have demonstrated that Pol ε is responsible for leading-strand DNA synthesis, while Pol α and Pol δ synthesize the lagging strand.^4–6 This gives rise to the canonical model of DNA replication, where synthesis progresses as part of a replication fork structure (Figure 1). Recently, our work in human cells demonstrated that the roles of Pol ε and Pol α are also conserved in human cells, establishing conserved roles of replicative polymerases between humans and yeasts.⁷ Generally, these roles are also confirmed in biochemically reconstituted replication forks.^8–10 Pol δ and Pol ε engage in highly accurate DNA synthesis, which is largely facilitated by strong base selectivity in their catalytic sites and proofreading action by their 3′–5′ exonuclease domains (reviewed in Bebenek and Ziuzia-Graczyk¹¹). Demonstrating the importance of the latter, mutations affecting the balance between polymerase and exonuclease activities of Pol ε cause a strong mutator phenotype in cancer cells. The contribution of such mutations to cancer is well documented and has been recently reviewed elsewhere.^12,13 The high fidelity exhibited by these replicative polymerases comes at the cost of flexibility, meaning that non-canonical templates are not tolerated and stall the progression of synthesis.

FIGURE 1. Roles of human DNA polymerases. During mitosis, canonical DNA replication is conducted at replication forks, where synthesis is performed by Pol α, Pol δ, and Pol ε following unwinding of DNA by the minichromosome maintenance (MCM) helicase. Damage in the template (star) has the potential to stall the progression of polymerases and, if left unresolved, result in the collapse of the fork. Polymerase stalling is resolved by one of several DNA damage tolerance (DDT) pathways; translesion synthesis (TLS), repriming, or fork reversal. Repriming produces gaps in the nascent DNA that must be filled by post-replication repair (PRR), using either TLS or homologous recombination (HR). Damaged bases within the DNA are be repaired by base excision repair, while strand breaks are repaired by HR, non-homologous end joining (NHEJ), or theta-mediated end joining (TMEJ). PCNA, proliferating cell nuclear antigen; Ub, ubiquitin.

Polymerases from other families that have roles in various DNA repair and DDT pathways are considered to be ‘specialized’ DNA polymerases and are typically error-prone. For example, members of the Y-family (Pol η, Pol κ, and Pol ι) play central roles in TLS, inserting nucleotides opposite damaged bases in the template strand (Figure 1). A unique B-family polymerase, Pol ζ, is thought to engage in extension from distorted termini left by other DNA damage-tolerant polymerases during TLS. While the synthesis activities of these enzymes on undamaged templates are prone to error, each TLS polymerase can accurately copy the genetic information from particular lesions, for example Pol η accurately bypasses ultraviolet-induced cyclobutane pyrimidine dimers.^14,15 Y-family polymerases are recruited to DNA via Rad18-dependent mono-ubiquitination of PCNA following DNA damage, facilitating a polymerase switch at sites of replicative polymerase stalling.^16,17 Recruitment may also be mediated by an interaction with another Y-family member, Rev1, which acts as a scaffold for TLS complexes.¹⁸

DNA damage bypass can also occur in a manner that is separated from the ongoing replication fork. In this process, following helicase-polymerase uncoupling, PrimPol synthesizes a nascent DNA primer downstream of a stalling obstacle to reprime synthesis (Figure 1, reviewed in Bainbridge et al.¹⁹). Efficient progression of replication forks can then resume, minimizing the exposure of ssDNA. Repriming events generate lesion-containing ssDNA gaps in the nascent DNA chain that must be filled behind the replication fork. Gap filling takes place by post-replication repair, whereby complementary DNA is synthesized either by TLS or template switching (reviewed in Gao et al.²⁰).

Polymerases from the A and X families generally have roles in DNA repair, except Pol γ, which engages in the replication of mitochondrial DNA. Several A and X family polymerases are involved in end-joining pathways, which ligate DSBs. For instance, Pol λ, Pol μ, and terminal deoxynucleotidyl transferase are reported to process broken strand ends and fill gaps during NHEJ (Figure 1).²¹ Pol θ, on the other hand, is identified as a key factor for an alternative mutagenic form of NHEJ known as TMEJ, where strands are joined based on the alignment of microhomologies in ssDNA at two broken ends (reviewed in Kruchinin and Makarova²²). Pol β from the X family plays an important role in the base excision repair pathway by filling gapped DNA and processing the end of the DNA backbone.²³ While A family polymerase Pol ν is conventionally ascribed a role in ICL repair, this is not supported by in vivo data, which instead suggest a role in meiotic homologous recombination.²⁴

MISREGULATION OF ERROR-PRONE POLYMERASE IN CANCER

Many factors have the potential to influence the fidelity of DNA replication. One mechanism that may increase mutagenesis during DNA synthesis is the imbalanced use of error-prone polymerases. Non-replicative DNA polymerases are believed to frequently function during replication, even in unperturbed conditions. For instance, while Pol ζ is the major cause of damage-induced mutagenesis in budding yeast it also underlies most spontaneous mutagenesis.²⁵ Recently, Pol ζ, in combination with Rev1, was shown to be responsible for a large proportion of spontaneous base substitutions in cultured human cells too.²⁶ These results can be interpreted as the involvement of mutagenic Pol ζ synthesis in unperturbed replication. Supporting a pivitol role for Pol ζ in completing replication, Pol ζ inactivation has severe consequences in mammalian cells, including the generation of DSBs and micronuclei.²⁷ Additionally, error-prone Pol η is proposed to compete with Pol δ for synthesis of the lagging strand, at the cost of fidelity.²⁸ Together, these findings suggest that non-replicative polymerases can participate in the replication machinery in a stochastic manner. An increase in the abundance of error-prone polymerases may therefore increase their usage during synthesis. Supporting this, biochemical studies using E. coli proteins have demonstrated that replicative and specialized polymerases compete for access to DNA in a concentration-dependent manner²⁹ and in the eukaryotic system replicative polymerases readily exchange within active replication machinery.³⁰ However, it is possible that a regulatory step, such as PCNA ubiquitylation, may be rate-limiting for the recruitment of non-replicative polymerases for synthesis (see Section 4.1). Ultimately, whether the abundance of error-prone polymerases influences their usage during replication has not been directly tested in eukaryotes. It is therefore useful to consider how altered levels of error-prone polymerases in cancer cells can influence the mutational load, cell phenotypes, and clinical progression. This section focuses particularly on Pol η, Pol κ, and Pol θ since many recent studies reveal roles for these enzymes in DNA replication during tumorigenesis. However, it is important to acknowledge that alterations to additional polymerases omitted here, such as Rev1,³¹ can also influence mutagenesis and carcinogenesis.

Pol η expression is commonly increased in cancers

Since the discovery of error-prone Y-family polymerases, levels of their transcription and abundance have been investigated in various types of cancer to explore their potential influence on tumorigenesis. Nowadays, the availability of cancer multi-omics data provides an overview of the general trends for each polymerase across many cancer types. Analysis of data from TCGA using cBioPortal³² shows Pol η mRNA is remarkably overexpressed in some cancers (Figure 2) due to frequent amplifications of the Pol η-encoding gene (POLH; Figure 3A). Interestingly, POLH expression increases in response to drug-induced replication stress in multiple cell lines, aiding recovery and survival.³³ Since replication stress is prevalent in cancer cells, the upregulation of Pol η observed in patient tumors may contribute to the fitness of cancer cells by mitigating such stress with its flexible synthesis activities. This notion is in line with the substantial excess of single and tandem somatic mutations observed in many cancers within sequence contexts known to be vulnerable to Pol η errors.³⁴ However, an experiment conducted in human fibroblasts found that the overproduction of Pol η does not increase spontaneous mutations,³⁵ suggesting that the abundance of Pol η alone is unlikely to be a causative factor for mutations in cancers. In many cancer cells, multiple regulatory mechanisms (such as the replication checkpoint and PCNA ubiquitylation) are also altered, which may activate or assist the functions of Pol η. It is therefore likely that a combination of regulatory mechanisms and the transcriptional status of Pol η are key determinants for Pol η usage and associated mutation load.

FIGURE 2. Expression pattern of polymerase encoding genes in various types of cancer. Thirty two tumor groups were queried using cBioPortal32 (www.cbioportal.org) and public datasets available as of December 2023. The heatmap color represents mRNA expression z-scores relative to adjacent normal samples.

FIGURE 3. Variation of Pol η-encoding gene (POLH) (A) or Pol κ-encoding gene (POLK) (B) expression and associated copy number variations and underlying structural changes in cancer samples. Thirty tumor groups were queried using cBioPortal32 (www.cbioportal.org) and public datasets available as of December 2023.

Variable functions of Pol κ depending on cancer stage or type

In contrast to Pol η, the expression of Pol κ mRNA is reduced in many cancers (Figure 2), often due to the deletion of the Pol κ-encoding gene (POLK; Figure 3B). This is consistent with a previous study that showed Pol κ expression was usually unaltered or decreased in the nine types of cancer examined.³⁶ Various phenotypes could arise on Pol κ down-regulation. For instance, a study using Xenopus egg extracts demonstrated that Pol κ plays a part in the regulation of the replication checkpoint by conducting short patches of DNA synthesis, generating the single-strand termini that facilitate the checkpoint factors.³⁷ In addition, Pol κ controls the stability and abundance of checkpoint kinase 1 (CHK1) and its downregulation may diminish the checkpoint response.³⁸ Therefore, a reduction of Pol κ may minimize checkpoint activity, which could be advantageous during the early stages of tumorigenesis to maintain the growth of cancer cells. Notably, TCGA data reveals that similar to Pol κ, mRNA levels of DNA polymerases involved in DDT (Pol ζ:REV3L, Pol ι, PrimPol) are also decreased in tumors (Figure 2). Therefore, a similar mechanism or selective pressure to downregulate these polymerases may also exist.

While it is not seen in the overall TCGA datasets, some studies report occasions where Pol κ expression is increased in cancers, particularly in gliomas, esophageal cancers, and lung cancers,^39–41 which can result from the loss of transcriptional repression by p53.⁴² This is potentially a major cause of the high mutation loads observed in these cancers since Pol κ is highly mutagenic, highlighted by dramatic increases in rates of mutagenesis on its overexpression^43,44 and also the observation of a Pol κ-specific mutagenic signature in unperturbed cultured cells.⁴⁵ In glioma, increased Pol κ expression was associated with advanced stages of the disease and identified as an independent prognostic marker,⁴⁰ suggesting that the requirement for Pol κ may differ throughout the progression of cancer. For instance, in later stages, when replication stress becomes more prevalent, cells with a higher abundance of Pol κ may be more tolerant of replication stress, encouraging their selection.

It is possible that Y-family polymerase expression fluctuates throughout tumorigenesis to assist tumors with altered conditions as they arise. For instance, in response to targeted clinical treatments that induce stress via oncogene inhibition, such as EGFR inhibition, expression of Pol κ and Pol ι mRNA is upregulated.⁴⁶ A concomitant downregulation of key accurate DNA repair pathways (mismatch repair and HR) also occurs. Together, this results in stress-induced mutagenesis, which encourages the acquisition of chemoresistance via a pathway reminiscent of bacterial adaptive mutability, which relies on the Pol κ homolog DinB polymerase.⁴⁷ Notably, expression of other Y-family polymerases can also confer resistance to treatment, for example p53-mediated upregulation of Pol η following treatment with various ICL-inducing genotoxic agents provides cells with a tolerance to such chemotherapeutics.⁴⁸

Multiple roles of Pol θ upregulation in tumorigenesis

The A-family polymerase Pol θ is drastically upregulated in some types of cancers (Figure 2) and numerous studies report a negative correlation between Pol θ expression and patient survival in a wide variety of cancer types (reviewed in Kruchinin and Makarova²²). With roles in both DSB repair and TLS, Pol θ is a highly mutagenetic polymerase, with comparable substitution mutation rates to the Y-family polymerases.⁴⁹ Accordingly, examinations of lung adenocarcinoma datasets from TCGA revealed an association between high Pol θ expression and an increased number of somatic mutations in tumors.⁵⁰ Interestingly, the abundance of Pol θ and other key TMEJ proteins (PARP1, Lig3, and Mre11) increases following the expression of oncogenic KRAS (KRAS^G12D) via a post-transcriptional mechanism and, as a result, mutagenic TMEJ is encouraged.⁵¹ Disrupting TMEJ by Pol θ deletion in mice expressing oncogenic KRAS delays the development of tumors, suggesting this pathway is a key player in KRAS-induced tumorigenesis. This is supported by data from TCGA which reveals that low Pol θ expression can increase survival in pancreatic adenocarcinoma patients with oncogenic KRAS mutations. Pol θ has additionally been implicated in protecting leukemia cells from DNA damage induced by metabolic changes.⁵² These metabolic changes increase cellular concentrations of formaldehyde, a potent crosslinking agent capable of covalently linking proteins to DNA. In response, Pol θ is upregulated to efficiently repair DNA–protein crosslink-induced DSBs by TMEJ. Overexpression of Pol θ therefore supports cells during tumor development and progression. On the other hand, repression of Pol θ expression can also support cancer progression in some instances. For example, transcriptional repression by epithelial-mesenchymal transition (EMT) transcription factor ZEB1, which regulates Pol θ expression via a direct interaction with its promoter,⁵³ reduces TMEJ, decreasing mutagenetic genetic alterations at the expense of overall genomic integrity. This potentially identifies a unique reliance of mesenchymal cells on canonical NHEJ.

ERROR-PRONE POLYMERASES ARE RECRUITED TO ALLEVIATE ONCOGENE-INDUCED REPLICATION STRESS Rad18-dependent PCNA ubiquitylation in cancer cells

A recent analysis of TCGA data did not detect an increase in mutation number in tumors with high Y-family polymerase expression in most of the 18 cancer types examined, despite their prevailing misregulation in cancer cells.⁵⁰ This may be indicative of an additional regulatory step governing the use of these enzymes. As the primary regulator of Y-family polymerases, PCNA ubiquitylation recruits these polymerases to DNA, promoting damage-bypass in many biological systems (Figure 1).^54,55 When replication is blocked by damaged DNA bases, the Rad6-Rad18 E2-E3 ligase complex is recruited to RPA-coated ssDNA,^56,57 where it mono-ubiquitylates PCNA to promote TLS. Subsequently, PCNA can be poly-ubiquitylated by the Ubc13-Mms2-Rad5 complex to initiate damage bypass by HR-dependent template switching. This process is relatively simple in the budding yeast system, in which Ubc13-Mms2-Rad5 catalyzes stepwise poly-ubiquitylation. However, the mechanism of PCNA polyubiquitylation is more complex with human proteins – the en bloc ubiquitin chain transfer⁵⁸ and the stepwise chain elongation are regulated differently (detailed in Masuda et al.⁵⁹).

Notably, PCNA ubiquitylation is also implicated in the replication of an unchallenged genome. Owing to the genetic interaction of Rad18 with replication components, Rad18-mediated PCNA ubiquitylation is correlated with the mechanism of lagging strand DNA synthesis. Accordingly, in fission yeast, in which PCNA is ubiquitylated during unperturbed growth, this pathway allows the efficient completion of Okazaki fragment synthesis by mediating gap filling.⁶⁰ Similarly, in 293T cells, PCNA ubiquitylation is shown to promote the filling of gaps which arise between Okazaki fragments, ultimately preventing degradation of the nascent strand.⁶¹ Consistently, an in vitro system using human proteins demonstrated that PCNA mono-ubiquitylation can be coupled to DNA replication per se and specialized polymerases also compete for access to the extending DNA terminus.⁶²

The accumulated evidence that links PCNA ubiquitylation with the process of DNA replication suggests that replication-driven PCNA ubiquitylation may be especially influential in proliferating cancer cells. Indeed, in many cancer-derived cell lines, PCNA ubiquitylation is induced during unperturbed S phase. Summarizing many studies focusing on specific cancers, expression of Rad18 is increased in a variety of cancer types, including melanoma,⁶³ esophageal cancer,⁶⁴ glioma,⁶⁵ colorectal cancer, gastric cancer,⁶⁶ and smoking-associated cancers.^67,68 Importantly, in these examples, high Rad18 expression generally correlates with tumor progression and recurrence and a poor prognosis, representing increased proliferation, invasiveness and chemoresistance.^63,64,66,69 Overexpression of Rad18 in cultured cells induces PCNA ubiquitylation and results in increased association of Y-family polymerases with chromatin.⁷⁰ Furthermore, Rad18 expression correlates with SNV burden in cancer datasets from TCGA.^67,68,71 This infers that increased Rad18 expression in tumors increases the use of Y-family polymerases, which characteristically introduce SNVs into nascent DNA due to their error-prone nature.

Adding a layer of complexity to their own regulation, some Y-family polymerases promote Rad18-mediated PCNA ubiquitylation. A non-catalytic function of Pol η facilitates efficient Rad18 foci formation and ensuing PCNA mono-ubiquitylation and Y-family polymerase recruitment.⁷² This phenomenon is mediated by two PCNA-interacting protein boxes located within the C-terminal region of Pol η.⁷³ Similar dual functionality was observed for the two PCNA-interacting protein boxes of Pol κ, which also promotes PCNA ubiquitylation, but not for Pol ι. Most intriguingly, PCNA ubiquitylation increases on Pol η overexpression in a dose-dependent manner,⁷² suggesting that altered expression of Pol η, and likely also Pol κ, in cancers could influence the regulation of polymerases during replication.

Rad18-mediated polymerase activation following oncogene expression

During the early stages of tumorigenesis, activation of oncogenes alters the metabolic landscape to promote proliferation and drive cancer progression. Oncogene expression is frequently accompanied by replication stress, a condition where the progression of replication is impeded or disturbed, leading to slowed or stalled forks (reviewed in Kotsantis et al.⁷⁴). Replication stress is tightly linked to the deregulation of replication mechanisms, which cells must tolerate in order to proliferate. Following the expression of the cyclin E or Ras oncogenes, an increase in Rad18-mediated mono-ubiquitylation of PCNA is observed.⁷⁵ This results in the specific recruitment of Pol κ to chromatin, where it maintains ongoing replication in the face of replication stress. In the absence of Pol κ, replication fork velocity is significantly decreased following cyclin E expression, indicative of fork stalling, and cells experience a dramatic increase in ssDNA gaps and DSBs. Interestingly, PrimPol-mediated repriming has recently been shown to aid cells with tolerating oncogene expression (KRAS^G12V) by maintaining ongoing DNA synthesis, highlighting a potential mechanism of ssDNA gap formation.⁷⁶ It may be that Pol κ has a role in filling such gaps to restore the DNA to its stable, double-stranded form. Mono-ubiquitylation of PCNA is also observed in response to Myc-induced replication stress caused by hyperactivation of origin firing, which results in the recruitment of Pol η to chromatin.⁷⁷ An interesting observation is that, although Pol κ and Pol η are both regulated by PCNA ubiquitylation, they are recruited in response to different oncogenes, which likely represents the distinctive functions of each polymerase being best suited to the specific challenges induced by each oncogene. The mechanism underpinning this selective recruitment is yet to be clearly defined.

While it is clear that oncogenic replication stress results in error-prone polymerase use via Rad18-mediated PCNA ubiquitylation, it remains unclear exactly how these polymerases contribute to alleviating replication stress, since oncogenes disrupt replication in many ways. Multiple studies examining cell lines ectopically expressing oncogenes identify dysregulated initiation of replication forks as a major cause of replication issues (reviewed in Hills and Diffley⁷⁸). For instance, ectopic expression of cyclin E, which is overexpressed in many cancers, can drive cell cycle progression and cause early entry into S-phase before a cell has prepared sufficient origins. In this situation, DNA synthesis begins with inadequate origins, which ultimately leads to under-replicated DNA due to impaired fork progression and reduced rates of synthesis. Other oncogenes, including cyclin E, human papillomavirus E6/E7, Myc, and Ras, are also suggested to disrupt the temporal regulation of replication initiation during the S phase and induce excessive origin firing. As a consequence, cells experience increased conflicts between replication and transcription machineries (cyclin E) and depleted dNTP substrates (E2F) due to overconsumption, both of which prevent efficient progression of polymerases and require resolution. As inherently flexible enzymes, error-prone polymerases are well-suited to alleviate these various consequences of oncogene-induced stress.

Whether these oncogene-induced replication problems occur genome-wide or only at specific genomic regions is currently unclear. Common fragile sites (CFS), which are sites of frequent structural rearrangement, undergo oncogene-induced destabilization. Importantly, Pol η is required to prevent under-replication at CFS (FRA7H and FRA16D in U2OS cells).⁷⁹ Our recent analysis of replicative polymerases revealed that leading-strand DNA synthesis by Pol ε is compromised at CFS (FRA3B and FRA16D in HCT116).⁷ Together, these results suggest that impediment of the replicative polymerase at CFS causes an alternative DNA polymerase, such as Pol η, to conduct synthesis. While the requirement of Rad18 for efficient replication at CFS has not been directly examined, it was recently reported that Rad18-mediated PCNA ubiquitylation is required to suppress R-loops and transcription-replication conflicts at genes within CFS.⁸⁰ Given this evidence, CFS likely represents an example of genomic loci that require flexible usage of DNA polymerases, mediated by Rad18, to tolerate oncogene-induced replicative disruptions.

It should also be noted that Rad18 expression may also exert detrimental effects on cancer cells via its involvement in other pathways. For instance, overexpression of Rad18 has been shown to induce the c-Jun N-terminal kinase-matrix metalloproteinase (JNK-MMP) pathway to encourage metastasis in esophageal squamous cell cancer,⁶⁴ increase cyclin D expression via activation of the phosphoinositide-3-kinase-Atk (pAtk) pathway in melanoma cells,⁶³ and mediate activation of the G2/M checkpoint to maintain genomic integrity.⁸¹ Taken together, it appears likely that a combination of these effects increases the overall severity of cancers with high Rad18 expression.

ALTERED POLYMERASE USAGE COMPENSATES FOR MUTATED PATHWAYS

Oncogenesis is frequently accompanied or even facilitated by mutations in genes from DNA repair or DDT pathways. Since cells possess significant redundancy between the different pathways that maintain genomic stability, the absence of a default pathway can be compensated by alternative pathways to overcome challenges. However, while the use of nonoptimal pathways can be sufficient to maintain proliferation, it is often erroneous. Since a number of recent studies have identified replicative defects in BRCA-deficient cells, this section utilizes this example to examine how polymerase usage can be modified to promote the survival of cancer cells.

PrimPol-dependent fork restart maintains gapped DNA synthesis in BRCA-deficient cells

Replication pathways are dramatically altered in cancer cells deficient in the hereditary breast cancer susceptibility genes BRCA1 or BRCA2, which are the most frequently mutated genes in familial breast and ovarian cancers. These two genes act as tumor suppressors by maintaining genome stability via roles in the fork reversal and HR pathways, both of which are disrupted upon the loss of BRCA proteins (reviewed in Gorodetska et al.⁸²). BRCA-deficient cells are hypersensitive to the chemotherapeutic agent cisplatin and PARP inhibitors. While HR deficiency was originally believed to underlie these sensitivities, an alternative model involving unrestrained replication fork progression has recently been proposed to better fit the observed phenotypes (Figure 4, described in detail elsewhere⁸³). This proposition followed the discovery that BRCA-deficient cells fail to restrict the progression of DNA replication in response to stressful conditions, such as treating cells with cisplatin or depleting nucleotide pools with hydroxyurea, which would normally stall DNA synthesis.⁸⁴ The DNA synthesized under these conditions was found to contain ssDNA gaps. Such gaps can also arise from defects in Okazaki fragment processing during inhibition of PARP, which traps the DNA break-detecting protein on DNA.⁸⁵ Furthermore, gaps in nascent DNA from BRCA-deficient cells were also observed in unperturbed cells, suggesting they form during a normal replication cycle in the absence of exogenous replication stress.^86,87 Since BRCA1 and BRCA2 usually inhibit expansion of gaps by limiting the activity of the MRE11 nuclease and their loss causes increased fork speed, BRCA-deficient cells are particularly vulnerable to gap accumulation.⁸⁸ Consequently, replication stress in these cells can induce the formation of excessive ssDNA gaps that decrease chromosomal stability. This ultimately induces apoptosis and underpins the chemosensitivity of BRCA-deficient cancer cells.⁸⁹

FIGURE 4. Consequences of altered DNA damage tolerance pathway availability. The response to fork stalling in the absence of breast cancer gene (BRCA) proteins. On encountering an obstacle (star), stalled forks can either be stabilized by fork reversal or restarted by repriming. In BRCA-deficient cells, reversed forks are unprotected and prone to degradation, therefore repriming is favored in these cells. Repriming permits unrestrained progression of synthesis but generates lesion-containing ssDNA gaps, which are filled by error-prone polymerases, facilitating the completion of replication. Incomplete gap filling produces persistent ssDNA gaps, at the detriment of cell fitness.

ssDNA gaps can arise as a result of a replisome skipping over an obstacle during synthesis, as occurs during repriming¹⁹ (Figure 4). Consistent with the increased presence of ssDNA gaps, expression of the human repriming enzyme PrimPol is elevated in BRCA1-deficient cells.⁸⁷ Importantly, ataxia telangiectasia and Rad3 related (ATR)-dependent upregulation of PrimPol has been shown to provide BRCA1-deficient cells with an adaptive tolerance to cisplatin, when administered in multiple doses, mimicking a clinical treatment regime.⁸⁶ This upregulation allows replication to progress in perturbed conditions, avoiding the use of fork reversal which is unstable in the absence of BRCA proteins, therefore enhancing cell proliferation and viability. PrimPol has also been shown to facilitate unrestrained fork progression in BRCA2-deficient cells in response to ionizing radiation, which is usually suppressed by an interaction between BRCA2 and MCM10.⁹⁰

Multiple error-prone polymerases fill ssDNA gaps in BRCA-deficient cells

As previously shown in budding yeast, in osteosarcoma-derived U2OS cells ssDNA gaps can be filled via two pathways; either TLS or HR.^88,91 However, in BRCA-deficient cells, the HR pathway is disrupted and therefore a reliance on TLS emerges.⁸⁷ Accordingly, recent investigations have specifically implicated the error-prone polymerases Pol θ and Rev1-Polζ in the filling of gaps that arise in BRCA1-deficient cells.^87,92

The use of these polymerases to fill ssDNA gaps has two major implications for cancer cells. First, the increased use of error-prone polymerases offers further opportunities for point mutations. It is well established that the loss of BRCA proteins causes elevated rates of mutagenesis.⁹³ Moreover, a recent analysis of the increased mutations in BRCA-deficient avian cells attributed the mutagenic signatures of base substitutions and short insertions/deletions to the use of error-prone Y-family polymerases, particularly Pol κ.⁴⁵ In human HCT116 cells, the use of Rev1-Polζ to fill PrimPol-mediated gaps in BRCA-deficient cells also increases rates of mutagenesis by 2- to 3-fold compared to BRCA-proficient cells.⁸⁷ Synthesis by Pol θ is inherently low-fidelity, but in the context of gap filling the enzyme produces deletion products in vitro via microhomology-mediated gap skipping, whereby the extending 3′ terminus is annealed by microhomologies within the ssDNA gap leading to the deletion of the intervening sequence.^49,92 In addition, synthesis by PrimPol is inaccurate, often misincorporating nucleotides and frequently introducing indels.⁹⁴ Thus, the increased use of these error-prone enzymes likely accelerates mutagenesis in BRCA-deficient tumor cells, especially in response to DNA lesions which block replicative polymerases.

Second, and perhaps most importantly, the use of TLS polymerases to alleviate the deleterious ssDNA gaps provides cancer cells with a mechanism to overcome chemosensitivity, oncogene-mediated senescence, and cell death, thereby providing resistance to replication stress.⁹⁵ In line with this, it has been demonstrated that upregulating TLS by disrupting either Fanconi anemia group J protein (FANCJ) or p21 can confer resistance to both chemotherapeutic- and oncogene-induced replication stress by encouraging gap suppression.⁸⁹ Importantly, such an acquired dependence on TLS enzymes has the potential to be leveraged in clinical applications. For instance, in the absence of BRCA1, the inhibition of Rev1 decreases the viability of cells and the growth of implanted tumors.⁸⁷ Likewise, inhibiting Pol θ in BRCA-deficient cells causes cell death in a synthetic lethal interaction.^92,96 Given this requirement of Pol θ in this particular genetic background, a Pol θ inhibitor is currently undergoing clinical trials to assess their efficacy for treating cancers with defective DNA repair pathways, either as a monotherapy or in combination with damage-inducing anti-cancer agents.

CONCLUDING REMARKS

It is clear from TCGA datasets that the expression of DNA polymerases, including those not covered in this review, is frequently misregulated in cancers (Figure 2). The extreme heterogeneity observed between different tumor types and likely between individual patients demonstrates the complex roles these enzymes play in tumorigenesis. While it was historically believed that error-prone polymerases primarily affect cancer progression via mutagenic synthesis, it is becoming increasingly clear that mutagenic synthesis represents only part of these enzyme's contribution to tumorigenesis. Recent evidence suggests that tumors are heavily dependent on specialized polymerases to tolerate disruption to replication pathways, and this is perhaps their primary contribution to tumorigenesis. These adaptations provide flexibility to overcome the disruptions to the dynamics of DNA replication that are frequently observed in cancer, albeit with compromised replication fidelity, allowing cancerous cells to survive and genomic instability to propagate. However, these recent discoveries are already beginning to unveil potential vulnerabilities in cancer cells of particular genetic backgrounds. Our thorough understanding of how cells alter the use of replication pathways in response to the disruption of individual pathways will therefore be crucial for the development of novel cancer therapeutics in the coming years.

AUTHOR CONTRIBUTIONS

Lewis J. Bainbridge: Conceptualization; investigation; writing – original draft; writing – review and editing. Yasukazu Daigaku: Conceptualization; data curation; investigation; supervision; writing – original draft; writing – review and editing.

ACKNOWLEDGEMENTS

The authors have nothing to report.

FUNDING INFORMATION

This work was supported by the JST FOREST program (JPMJFR204X, Y.D.), JSPS KAKENHI grants (JP23H02463) to Y.D. and the Naito Foundation, Takeda Science Foundation and Yamada Science Foundation (Y.D.).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

ETHICS STATEMENTS

Approval of the research protocol by an Institutional Reviewer Board: N/A.

Informed Consent: N/A.

Registry and the Registration No. of the study/trial: N/A.

Animal Studies: N/A.