Structure and function relationships in mammalian DNA polymerases

Abstract

DNA polymerases are vital for the synthesis of new DNA strands. Since the discovery of DNA polymerase I in Escherichia coli, a diverse library of mammalian DNA polymerases involved in DNA replication, DNA repair, antibody generation, and cell checkpoint signaling has emerged. While the unique functions of these DNA polymerases are differentiated by their association with accessory factors and/or the presence of distinctive catalytic domains, atomic resolution structures of DNA polymerases in complex with their DNA substrates have revealed mechanistic subtleties that contribute to their specialization. In this review, the structure and function of all 15 mammalian DNA polymerases from families B, Y, X, and A will be reviewed and discussed with special emphasis on the insights gleaned from recently published atomic resolution structures.

Introduction

Without a way to replicate the genome, life as we know it would not be possible. DNA polymerases have evolved to fill this role, all of which perform a similar chemical reaction: the addition of a nucleotide monophosphate (dNMP) onto the end of a DNA primer sequence [1, 2]. As novel DNA polymerases are discovered and characterized, this nucleotidyl transferase reaction of DNA polymerization has been revealed to occur in diverse cellular situations aside from genome replication. These situations include the replication of mitochondrial DNA, synthesis from diverse substrates generated by DNA damage and during DNA repair, and error-rich DNA polymerization to provide antibody diversification [3, 4]. This cellular need for diverse types of DNA synthesis has led to the evolution of many specialized polymerases. Currently, 15 independent mammalian DNA polymerases have been characterized, each of which fulfills a specific biological role (Table 1) [3]. Based on sequence homology, mammalian DNA polymerases can be classified into four families: A, B, X, and Y [5, 6]. Generally, A-family DNA polymerases are involved in DNA lesion bypass and mitochondrial DNA replication and repair, B-family DNA polymerases are largely involved in genomic DNA replication, X-family DNA polymerases are involved in DNA repair and V(D)J recombination, and Y-family DNA polymerases are mainly involved in lesion bypass [7]. Each mammalian DNA polymerase will be discussed in detail in this review, with special emphasis given to the structural adaptations used to fill its biological niche.

Table 1.

Selected features of human DNA polymerases [4, 7, 40–42, 64, 67, 83, 84, 95, 114, 121, 135, 136, 259, 282, 288]

Family	Pol	Catalytic subunit mass(kDa)	Structure without DNAa	DNA bound structurea	Fidelityb	Additional subunits	Biological pathwayc
B	Alpha (α)	165	4B08	4Q5V, 5IUD	10−4	Subunit B, primase large, and primase small subunit	DNA replication
	Delta (δ)	100		3IAY	10−5	PolD2, PolD3, and PolD4	DNA replication
	Epsilon (ε)	225		4M8O, 4PTF	10−5	p59, p12, and p17	DNA replication
	Zeta (ζ)	353	3ABD		10−3 f	Rev7, PolD2, and PolD3	TLS
Y	Eta (η)	78		3MFH	10−2	Monomer	TLS
	Iota (ι)	80		1T3N, 2FLN, 4EYH	10−1	Monomer	TLS
	Kappa (κ)	76	1T94	2OH2, 4U6P, 6CST	10−3	Monomer	TLS
	REV1	138		2AQ4, 5WM8, 3BJY	N.A.d	Monomer	TLS
X	Beta (β)	39	1BPB, 1RPL	2BPG, 2FMS, 31SB	10−4	Monomer	BER
	Lambda (λ)	66	5CB1	1RZT, 2BCQ	10−3	Monomer	BER, NHEJ, TLS
	Mu (μ)	55	4LZD	2IHM, 4LZG	10−3	Monomer	NHEJ, V(D)J
	TdT	56	1JMS	4I27	N.D.e	Monomer	V(D)J recombination
A	Gamma (γ)	140	3IKM	4ZTZ, 5C53	10−5	Two copies of subunit 2	Mitochondrial DNA replication and repair
	Theta (θ)	290		4X0P	10−3	Monomer	TLS, MMEJ
	Nu (v)	100		4XVK	10−3	Monomer	ICL repair

^aRepresentative crystal structures discussed in this review

^bApproximate error rate for single-base substitutions

^cThis column focuses on the primary biological pathway(s) of each polymerase; others may also apply

^dFidelity measurements are not applicable to REV1

^eNo data could be found for the fidelity of TdT’s templated activity, although it is likely similar to other X family polymerases

^fMeasurement from yeast homolog

The first detailed information regarding key amino acids necessary for DNA polymerase activity became available in the mid-1980s through structural studies of Escherichia coli DNA polymerase I [8]. Further advances, such as the crystallization of polymerases bound to oligonucleotides and time-lapse X-ray crystallography, have provided insight into DNA polymerase function in context of their DNA substrates [9]. As additional structures were published, including proteins from the Eukarya domain, it became increasingly clear that the catalytic subunits of these enzymes exhibit a common overall architecture [8, 10, 11]. This architecture has been likened to the structure of a human right hand, containing a palm, fingers, and thumb domain [12, 13]. When complexed with a DNA substrate, this hand can be envisioned to have DNA threaded through it, with the fingers binding the incoming nucleotide and coordinating the single-stranded template. The palm contains the catalytic core, which interacts with both the incoming nucleotide and dsDNA, while the thumb “grips” the dsDNA upstream of the polymerization site [13]. Many DNA polymerases exhibit global conformational changes upon the binding of a nucleotide, moving from an open to a closed state prior to catalysis. Continuing the hand analogy, this is accomplished by a shift in the fingers domain, in the same manner that a hand would transition from an open to closed position upon grabbing a rope [14–16].

Because the catalytic subunit of each DNA polymerase exhibits a similar structure, it should not be surprising that the chemical reaction that occurs is nearly identical in the case of each polymerase. In this reaction, DNA polymerases add a free dNMP to the end of a primer, creating a phosphodiester linkage [1, 2]. This nucleotide addition utilizes the 3ʹ hydroxyl group (3ʹ-OH) at the end of the primer as a nucleophile in an S_N2 reaction, attacking the α phosphorous of the phosphate group of the incoming nucleotide triphosphate (dNTP) to form an incorporated nucleotide monophosphate and free pyrophosphate [2]. The conservation of this reaction between the mammalian DNA polymerases manifests itself in a generally universal mechanism. First, the DNA polymerase binds its DNA substrate, which usually consists of either gapped double-stranded DNA (dsDNA), or single-stranded DNA (ssDNA) with an annealed primer; in either case, the polymerase must access an exposed 3ʹ-OH. Then, a dNTP binds the complex (which is canonically complementary to the templating base along its Watson–Crick face), where it is poised for catalysis. Finally, catalysis occurs, during which the 3ʹ-OH is activated by a divalent cation (e.g., Mg²⁺) to form a phosphodiester bond between the primer terminus and the incoming dNMP, allowing the resultant pyrophosphate product to be released. Catalysis involves at least two divalent metals. A third metal has been observed in numerous crystallographic studies and the exact mechanistic role of the third metal ion remains an active area of research [17, 18]. DNA polymerase catalysis was visualized with pol β, η, and µ to study DNA synthesis via time-lapse X-ray crystallography, traversing from the ground state to the product state to observe phosphodiester bond formation [17, 19, 20]. As the number of published atomic resolution structures of DNA polymerase complexes increases, subtle, but still vitally important, structural specializations have been observed. Currently, all 15 mammalian DNA polymerases have published atomic-level structural information; the way these structures dictate their specialized functions will serve as a focal point of this review. See Table 1 for a summary of these DNA polymerases, their functions, and the families to which they belong.

B-family

During genomic replication, the six billion nucleotides that compose the human genome must be copied in a short amount of time, while minimizing the introduction of damaged and mismatched nucleotides [21]. Depending on their locations, erroneous base insertions may cause devastating cellular outcomes, including cell death, senescence, or carcinogenesis [22]. The monumental task of replicating the genome is performed by the B-family polymerases: DNA polymerases alpha (pol α), delta (pol δ), and epsilon (pol ε) (Fig. 1a). DNA polymerase zeta (pol ζ) is also a member of this family based on sequence homology, but is not responsible for bulk replication of the genome. Pol α acts as the initiator of DNA synthesis, using two separate active sites for dual activities, with its primase activity creating RNA primers on ssDNA and its polymerase activity extending them [23]. Pol δ and pol ε further extend those DNA primers, performing the bulk of DNA synthesis, and while controversial, many lines of evidence support pol ε replicating the leading strand and pol δ on the lagging strand [24]. Pol ζ, although homologous to the other members in the family, is not part of the traditional replisome, but assists when replication is stalled by DNA damage, such as during translesion synthesis, crosslink repair, and some forms of homologous recombination [25]. To ensure accurate replication of the genome, B-family polymerases have evolved to exhibit some of the highest levels of both processivity (the number of nucleotides inserted before the polymerase disassociates) and fidelity (the selection of the correct nucleotide to insert) [26]. The presence of a 3ʹ to 5ʹ exonuclease activity for proofreading improves fidelity by a factor of up to 100, shifting the error rate from around one misinsertion per 10⁵ insertions to one misinsertion per 10⁷ insertion events in pol δ and pol ε [27, 28]. The high processivity of B-family polymerases can be facilitated by interactions with proliferating cell nuclear antigen (PCNA), which acts like a clamp to keep the polymerases on the DNA [29, 30]. Working in concert, B-family polymerases compose a vital part of the DNA replication complex.

Fig. 1.

Overall structure and specific features of replicative B-family polymerases. a Surface representations of human pol α (5IUD), yeast pol δ (3IAY) and yeast pol ɛ (4M8O) are shown with their exonuclease (orange), P-domain (yellow), fingers (cyan), thumb (green), palm (blue), and N-terminal (red) domains highlighted. b A close-up view of the pol α thumb domain (green) making contact with the 2′-OH groups of the RNA of its DNA/RNA hybrid substrate (white sticks), the product of its primase activity (4Q5V). c The minor groove of nascent DNA probed by Pol δ through direct contacts (blue sticks), as well as through coordinated water-bridges (cyan spheres; 3IAY). d The P-domain (yellow), found only in Pol ɛ, provides additional contact to the nascent DNA (gray surface; 4M8O)

DNA polymerase alpha (pol α)

The B-family member DNA polymerase alpha (pol α) is involved in both DNA replication and cell-cycle regulation [31]. As the Greek letter α implies, pol α was one of the first DNA polymerases discovered and characterized [32]. It plays the crucial role of creating and extending primers at the beginning of genomic replication; therefore, pol α is tightly synchronized with the progression of the cell-cycle [33]. Pol α is made up of four subunits, the catalytic subunit, subunit B, DNA primase small subunit, and DNA primase large subunit (also named p180, p70, p49, and p58, respectively) [34]. Subunit B is a regulatory subunit, and the primase subunits together coordinate primase activity [35, 36]. Furthermore, solution studies of the primase subunits showed that both the large and small primase subunits are required for the primase activity [37]. Compared to other DNA polymerases, the catalytic subunit of pol α exhibits a moderate fidelity of nucleotide insertion, with an error rate of one base misinsertion per 1000 insertion events [7]. Although pol α contains an exonuclease domain, no proofreading activity has been observed; errors must, therefore, be corrected by other repair systems (Fig. 1a) [26, 38, 39].

The atomic resolution structure of the pol α catalytic subunit was initially obtained from a yeast homolog, which provided some of the first mechanistic insight [40]. Structures are now available of the human pol α catalytic subunit in multiple physiologically relevant ligand-bound states, including bound to a DNA:RNA hybrid and dsDNA helix substrates, which would be involved in DNA priming and DNA synthesis, respectively. Interestingly, the DNA in contact with pol α in the pol α:DNA complex structure is not in B form, but rather a hybrid A-B form, suggesting DNA sculpting by the enzyme during its polymerase activity [41]. Additionally, a structure of the catalytic subunit bound to its DNA:RNA substrate demonstrated that pol α directly coordinates certain 2ʹ-OH groups of the RNA primer, a unique substrate for pol α (Fig. 1b) [42]. Furthermore, this structure revealed the molecular contacts mediating inhibition by aphidicolin, a B-family polymerase inhibitor, which was shown to occlude the binding of the incoming nucleotide [42]. Altogether, these structures have acted to reveal the polymerase mechanism of pol α in both its native state and under conditions of perturbation via small molecule inhibitors.

Atomic resolution structures of the other three pol α subunits have also been obtained. The primase subunit structures were solved in complex with portions of both the B subunit and a majority of the catalytic subunit. This work provided particular insight on the conformational changes required for pol α to consistently create short RNA primers with its primase domain before transitioning to its polymerase domain. Specifically, these structures revealed an interdomain steric occlusion of the primase domains which limits primer elongation to ~ 10 ribonucleotides, as well as a global shift that allows the channeling of the primed DNA from the primase to the polymerase domain without dissociation [43]. Additionally, crystal structures of the large primase subunit revealed that an evolutionarily conserved iron-sulfur cluster is present in the protein’s core, an unusual finding as iron-sulfur clusters are almost exclusively found near the exterior of proteins [44]. Further investigation of the iron-sulfur cluster showed that the initiation of DNA synthesis may be triggered by the charge state of the iron-sulfur cluster in the primase domain, in which an electron is transferred from the polymerase domain to the primase domain via the DNA substrate. This altered charge state modifies the binding affinity and triggers the interdomain transport of DNA [45]. These structural advances in understanding all four subunits of pol α have been instrumental for gaining understanding of how DNA replication is initiated.

DNA polymerase delta (pol δ)

DNA polymerase delta (pol δ) is one of the major replicative DNA polymerases in eukaryotes and, in addition to participating in cell cycle checkpoints and multiple DNA repair pathways, is responsible for replication of the genome along with pol α and ε [46–49]. While the catalytic subunit of pol δ (POLD1) contains both the polymerase active site and exonuclease active site, the fully assembled pol δ is a heterotetramer comprised of POLD1 and three accessory subunits POLD2, POLD3, and POLD4 (also known as p125, p50, p66/p68, and p12, respectively) [50, 51]. These accessory subunits mediate many of the protein–protein interactions between pol δ and accessory proteins such as the clamp loader replication factor C (RFC) and PCNA [52–54]. When these binding regions are disrupted, or if PCNA is absent, processivity declines drastically, highlighting the critical role of PCNA in genomic replication [54]. Genomic replication is an exceedingly intricate process, which has made pol δ’s specific role in the replisome challenging to identify, and controversial to date [55–59]. Currently, one widely accepted model designates pol δ as responsible for the replication of the lagging strand of the replication fork, while pol ε concurrently replicates the leading strand [60, 61]. Conversely, some groups suggest pol δ may be the major polymerase on both strands with pol ε playing a less active role [56, 62]. In both models, however, pol δ replicates the lagging strand of the replication fork. Therefore, in either case, pol δ is an essential protein in eukaryotes via its vital role in genomic replication [63].

Because pol δ replicates billions of bases with each cell division, the mutational impact of misinsertion is amplified greater than other DNA polymerases that replicate on smaller scales. Accordingly, the fidelity of pol δ has evolved to be one of the highest among DNA polymerases, at a rate of ~ 1 base misinsertion per 10⁵ nucleotides replicated [64, 65]. Without a functional proofreading domain, the fidelity of pol δ decreases 60-fold, but still retains a moderately high fidelity from its tight active site pocket [65, 66]. This active site pocket was structurally characterized in 2009 when a ternary structure was reported of Pol3, the yeast homolog to the human pol δ catalytic subunit, in complex with a DNA template-primer and an incoming dCTP [67]. The structure revealed that upon dNTP binding, a rotation of the fingers domain promoted correct insertion by shaping a tight pocket around the incoming nucleotide [67, 68]. The understanding of pol δ’s fidelity also grew from the aforementioned structure, because it includes the exonuclease domain of Pol3. In other polymerases with exonuclease domains, extensive interactions on the minor groove side of the DNA were shown to probe the base pairing properties of nucleosides up to two bases anteceding the next incoming base [69]. Disruption of these interactions by a mismatch is thought to trigger a conformational change to transfer the substrate to the exonuclease domain [70]. In the case of Pol3, these extensive minor groove interactions continued as far as 5 base pairs down the DNA, with many interactions mediated by coordinated waters (Fig. 1c). Therefore, pol δ would have up to five insertion events to “catch” the mistake, leading to a more sensitive proofreading potential than many other DNA polymerases [67].

DNA polymerase epsilon (pol ε)

When DNA polymerase epsilon (pol ε) was first discovered in 1986, it was initially thought to be a high molecular weight form of pol δ [71]. However, in 1990 it was determined to be a distinct polymerase that functions in concert with pols α and δ to replicate the genome [72]. Even now, after over 20 years of research into this enzyme, the exact role of pol ε in the replicative process is still debated [73]. Studies utilizing the specific error signatures of mutant yeast pol δ and pol ε, as well as the observation of pol ε specific incorporation of ribonucleotides, describe a model where pol ε and pol δ synthesize the leading and lagging strand, respectively. Still, alternative models have been put forward and have been the subject of multiple reviews [56, 58, 60, 74, 75]. DNA pol ε has similar features to the other B-family polymerases, yet unique domains and features which confer the greatest fidelity and processivity of any known mammalian DNA polymerase [76–78]. As a B-family polymerase, human pol ε is a heterotetramer composed of a large catalytic subunit encoded by POLE1 and three accessory subunits: POLE2, POLE3, and POLE4 (which encode proteins p261, p59, p12, and p17/CHRAC-17, respectively) [79]. A 20 Å cryo-EM structure of the yeast pol ε heterotetramer revealed that these accessory subunits extend from the catalytic subunit like a tail, interacting with the DNA helix, and are thought to increase processivity through direct interactions with DNA processivity factors. Additionally, structures of the human homologs of some of these subunits are also available [80–82].

Although a high-resolution structure of the human catalytic subunit has yet to be solved, ternary structures (with DNA substrate and incoming nucleotide) of the yeast homolog were determined [83, 84]. Similar to other high fidelity polymerases, the fingers domain closes upon nucleotide binding to form a tight pocket around the inserted nucleotide to both position the dNTP for catalysis and reduce the frequency of mismatch insertions. Interestingly, this pol ε pocket includes a unique residue in the major groove (Tyr481), not conserved in pol δ, possibly contributing to its superior fidelity and potentially reducing the insertion efficiency of modified nucleotides that are handled efficiently by other DNA polymerases [85, 86]. This structure also gave the first glimpse of an additional domain on pol ε not present in other members of the B-family, the P-domain (Fig. 1a, d). This domain was observed to wrap around the DNA and is named for its impact on processivity [83]. Further assaying this processivity, in the absence of PCNA, pol ε was observed to insert up to 27 nucleotides prior to dissociation, compared to pol δ’s three. Additionally, both the processivity and activity of pol ε were stunted with the removal of the P-domain, highlighting its essential role within pol ε’s function [83].

Similar to other DNA polymerases with exonuclease domains, residues along the minor groove survey the newly synthesized primer strand through extensive interactions. Upon disruption by a mismatched base pair, the exonuclease process is triggered [67, 69, 83]. A nearly identical system is seen in pol δ (Fig. 1c); however, pol ε contains an additional hydrogen bond from a coordinated water molecule to the nucleoside at the primer terminus (not observed in pol δ) which may aid in mismatch recognition [83]. When a mismatch is recognized, a conformational shift in the thumb domain guides the primer strand ~ 41 Å from the polymerase active site to the exonuclease domain where the exonuclease reaction occurs, similar to pol δ. Thus, structures of the yeast homolog of pol ε have revealed how the stringent polymerase active site and exquisitely sensitive exonuclease domain give rise to its nearly unmatched fidelity.

DNA polymerase zeta (pol ζ)

In the 1970s, REV1, REV3, and REV7 were identified as genes associated with a deficiency in mutagenesis [87–90]. Homology of REV3 to other genes encoding B-family DNA polymerases, as well as its role in mutagenesis, led to the hypothesis that REV3 encodes a lesion-associated DNA polymerase [91]. This was confirmed in 1996 when the yeast Rev3 and Rev7 proteins were purified as a heterodimeric complex and characterized as the sixth eukaryotic polymerase, DNA polymerase zeta (pol ζ) [92]. As the only error-prone member of the B-family, pol ζ has a significantly lower fidelity compared to other replicative polymerases [93]. However, pol ζ is an important part of DNA damage tolerance, in coordination with the Y-family polymerases which specialize in translesion DNA synthesis (TLS). As the first DNA polymerase found to specialize in TLS, its discovery and characterization led to a new era in the polymerase field as focus shifted to enzymes specializing in translesion DNA synthesis [94].

In 2010, the first published crystal structure of the multi-subunit pol ζ complex revealed a portion of the human Rev3 catalytic subunit in complex with Rev7 (the smaller, non-catalytic subunit), mapping their interaction to a region N-terminal of the polymerase domain [95]. Rev7 stimulates the activity of the catalytic Rev3 subunit by 20–30 fold in vitro [92] and mediates several protein–protein interactions with additional proteins, suggesting many potential roles of pol ζ in diverse cellular processes [95–99]. Although Rev3 and Rev7 are sufficient for pol ζ activity in vitro [92], a stable four-subunit heterotetramer complex can be purified with accessory subunits PolD2 and PolD3 from pol δ. This pol ζ holoenzyme (composed of Rev3, Rev7, PolD2, and PolD3) exhibits increased activity and enhanced processivity compared to the heterodimeric Rev3–Rev7 complex [100, 101]. Organization of this entire pol ζ heterotetramer was observed via electron microscopy (EM), generating a 23 Å resolution bilobal 3D model of yeast pol ζ [102]. The large lobe can be fit with the Rev3 catalytic subunit, and is connected to the smaller lobe which consists of the Rev3 C-terminal domain, as well as the regulatory subunits Rev7, pol31, and pol32 (yeast homologs of PolD2 and PolD3). Two stems of density connecting the lobes suggest dual linkers that may result in a more rigid association than seen in the more flexible assemblies of pols α and δ, which have a similar elongated bilobal structure [103, 104].

Pol ζ is specialized for the extension step of lesion bypass and can readily extend from matched, mismatched, and damaged primer-termini [105, 106]. Therefore, two polymerases act sequentially, with pol ζ facilitating extension after another TLS polymerase bypasses the lesion. Pol ζ is also recruited to sites of replication fork stalling caused by defective replicative polymerases [107]. Such compromised replisomes stall more frequently at DNA lesions, activating error-prone bypass and extension of mismatched primers by pol ζ. The activity of pol ζ is important for sustaining chromosome stability, but like most TLS mechanisms, occurs at the cost of increased mutagenesis. Because of this, pol ζ has been implicated in a large portion of cellular mutagenesis [108, 109]. While partial crystal structures and EM models provide some insight into its role in DNA damage bypass and extension, minimal structural information is available for pol ζ in the context of damage; therefore, the structural basis for lesion bypass and extension is unclear at this time.

Y-family

Despite the numerous pathways that work to repair DNA, some lesions can persist and block subsequent rounds of replication, as replicative polymerases are unable to accommodate most DNA lesions within their compact active sites [64, 66]. As a result, damage tolerance pathways have evolved to ensure that replication can be completed in the presence of DNA damage. Y-family DNA polymerases are specialized to ensure continuity at the replication fork via incorporation of a nucleotide opposite the lesion by TLS. When a replicative polymerase encounters a lesion, a polymerase switching event occurs, replacing the arrested polymerase with a TLS polymerase specialized for lesion bypass. The first genes encoding translesion polymerases were identified in 1971 by Jeffery Lemontt via a damage-induced mutagenesis study [87]. Due to their almost complete lack of primary sequence homology with replicative DNA polymerases, it took the discovery and characterization of several more TLS polymerases before they were connected as a family in 2001 [110].

The four main TLS polymerases (which include DNA polymerases eta (pol η), iota (pol ι), kappa (pol κ), and Rev1) belong to the Y-family of DNA polymerases and are uniquely adapted to bypass different DNA damage lesions. Crystal structures have provided insight into structurally distinct features that convey their diverse strategies for replicating through lesions. In addition to the canonical polymerase ‘right-hand’ domains described in the introduction, TLS polymerases contain an extra polymerase-associated domain (PAD) (sometimes called little finger [111] or wrist domain [112]) that extends from the C-terminus and makes extra DNA stabilizing contacts [113]. Another feature that sets apart these damage-associated polymerases is the absence of a 3ʹ–5ʹ exonuclease domain that proofreads newly synthesized DNA. While most high-fidelity replicative polymerases have this domain, most of the specialized polymerases do not and thus exhibit lower fidelity [114]. Here, we will further structurally analyze the unique mechanistic strategies used by each Y-family polymerase to allow the bypass of specific lesions.

DNA polymerase eta (pol η)

Originally identified in 1996 by sequence homology with prokaryotic TLS polymerases [115, 116], the yeast RAD30 gene was shown to encode DNA polymerase eta (pol η) in 1999 [117]. It was not long until the human homolog (POLH) was identified and characterized, making pol η one of the founding members of the Y-family of DNA polymerases [110]. The predominant biological role of pol η is replication through cyclobutane pyrimidine dimers (CPDs) [117] and 8-oxoguanine [118]. CPDs are a type of DNA lesion resulting from the UV-radiation catalyzed covalent linkage of two adjacent pyrimidines. While CPDs block most DNA polymerases, pol η is specialized to efficiently and correctly insert adenines (A) opposite each thymine (T) in the dimer, resulting in error-free bypass of the lesion (Fig. 2a). In humans, defects in pol η activity due to POLH mutations lead to the variant form of Xeroderma Pigmentosum (XP-V), a genetic disease associated with increased risk of sunlight-induced skin carcinomas [119, 120].

Fig. 2.

Y-family polymerases. Unique lesion bypass mechanisms of a templating lesion and incoming nucleotide are indicated (white sticks). a human Pol η shown in surface (green; 3MR3) b human Pol ι shown in cartoon (green; 2DPJ) c human Pol κ shown in surface with indicated catalytic core (green), PAD (yellow), and N-clasp (cyan) (4U7C) d yeast Rev1 shown in cartoon with undamaged templating guanine (green; 2AQ4), overlayed with damaged guanine template (white) and residues (cyan sticks; 5WM1)

Crystal structures have shown that, compared to other polymerases, the pol η fingers domain is positioned further from the templating base, enlarging the active site [121]. Importantly, this particularly wide active site is spacious enough to accommodate two bases of undamaged DNA or both bases of a CPD lesion (Fig. 2a) [121, 122]. It was hypothesized that since the structural constraints of the replicative polymerases are absent, pol η relies on the stability of Watson–Crick hydrogen bonding to determine which bases are correct for subsequent incorporation. Supporting this hypothesis, studies with yeast pol η used nucleotide base analogs to demonstrate that pol η is more dependent on Watson–Crick hydrogen bonding between the templating and incoming nucleotide than other polymerases [123]. This explains why pol η can replicate through some lesions (such as CPDs, 6–4 TC or CC photoproducts [124], and 8-oxoguanine) but is inhibited by lesions that largely disrupt Watson–Crick base pairing [125].

Since CPDs are known to confer bending of the double helix, it is surprising that crystal structures have shown that CPD-pol η complexes retain a straight B-form DNA molecule. This phenomenon is attributed to the close interaction between the pol η catalytic core and PAD domain, which creates a positively charged pocket for DNA binding. Described as a “molecular splint”, this pocket holds the DNA in a conformation that allows the bypass of CPD lesions [122]. After the incorporation of adenines across from the CPD, pol η processively inserts at least two additional bases past the lesion before steric hindrance and the loss of hydrogen bonding causes it to dissociate. The position-dependence of CPD-pol η interactions produce a structural basis for polymerase switching after lesion bypass [122]. While this discussion focused on the primary role of pol η in TLS, the polymerase has also been implicated in multiple other processes such as somatic hypermutation [126, 127] and homologous recombination [128, 129].

DNA polymerase iota (pol ι)

Human DNA polymerase iota (pol ι), a Y-family polymerase encoded by the POLI (RAD30B) gene, was shown to have TLS activity in 2000 [130]. While its existence has been suggested to be the result of POLH (pol η) gene duplication, pol ι has notably different biochemical properties that convey a unique mechanism for lesion bypass [131, 132]. Polymerases typically rely on the well-defined geometry of Watson–Crick base pair interactions, resulting in about the same catalytic efficiency for correct insertion for all four bases. Even though pol ι is often referred to as one of the most error-prone DNA polymerases, its activity is highly sequence context dependent, having higher efficiency and fidelity for nucleotide incorporation opposite template purines than opposite template pyrimidines [130, 133, 134].

Crystal structures of pol ι ternary complexes (with DNA substrate and incoming nucleotide), reveal features of the active site that provide insight into the insertion and base-pairing preferences elucidated from kinetic studies. Unlike replicative polymerases, a conformational change is seen in the DNA upon pol ι binding, while the enzyme and active site remains rigid [135]. Upon dNTP binding, the template base is forced to rotate about its glycosidic bond into syn-conformation, facilitating Hoogsteen base pairing with the Watson–Crick face of an incoming dNTP (anti). This mechanism relies on the narrow active site of pol ι to impose the required syn-conformation of the templating base. Several aliphatic active site residues (Leu62, Val64, and Gln59) impose constraints on the template, effectively preventing Watson–Crick base pairing. In addition, the syn-conformation is stabilized by other protein contacts and a small hydrophobic cavity which directly positions the sugar moiety of the template [136]. Holding the template securely in the syn-conformation effectively limits its hydrogen bonding opportunities to bases with Hoogsteen hydrogen bonding capability, resulting in low efficiency, low fidelity, and inaccurate lesion bypass [130, 134].

Structures of pol ι complexes with damaged templating DNA reveal that this Hoogsteen base pairing mechanism allows for the bypass of various lesions, specifically ones that impair Watson–Crick interactions [137, 138]. For example, N²-ethylguanine [139], O⁶-methylguanine [140], and 1,N⁶-ethenoadenine [141] are damaged bases that adopt the syn-conformation in the active site of pol ι to allow Hoogsteen base pairing (Fig. 2b). For minor groove adducts, rotation of the nucleotides into the syn-confirmation removes the damaged moiety from the active site by repositioning it into the major groove of the dsDNA, where it can be more readily accommodated. Once removed from the active site, these lesions no longer interfere with the interactions between the template base, polymerase, and incoming nucleotide, which allows for efficient lesion bypass [138]. While structures have shown pol ι using its Hoogsteen face via this mechanism, other studies suggest it can alternatively use its Watson–Crick face [142, 143], reinforcing its diverse TLS function in the bypass of many forms of lesions.

DNA polymerase kappa (pol κ)

Like other TLS polymerases, the POLK gene was discovered via homology searches for eukaryotic orthologs of the E. coli dinB gene [144]. The POLK gene was cloned in 1999 and its gene product, DNA polymerase kappa (pol κ), was demonstrated to have TLS activity in 2000 [145, 146]. The TLS function of pol κ remains somewhat unclear, with two main proposed cellular roles: (1) extension of primer termini, and (2) the incorporation of nucleotides to bypass bulky minor groove adducts [138]. While in vitro studies support both functions, many studies point to extension as its main cellular role. Although most polymerases are inefficient at extension from mismatched primer-termini, pol κ extends from mismatched [147] and lesion-containing termini [148–151] with comparable efficiency as non-damaged matched termini [146]. Mismatched termini could often result from lesion bypass by other TLS polymerases, suggesting pol κ may have an important cellular role as an extender of mismatched and lesion-containing termini resulting from TLS. Additionally, a recent publication identified pol κ as having a role in promoting DNA synthesis and replication stress recovery at sites of stalled forks [152]. Thus, the functional role of pol κ is an active area of interest in the field.

When a crystal structure of the pol κ catalytic core without DNA present was published in 2004, it was the first structure of any human Y-family polymerase [153]. Subsequently, in 2007, the structure of a pol κ-DNA ternary complex with an incoming nucleotide provided additional insight in the context of DNA. Specifically, it revealed the importance of a unique 100-residue N-terminal extension, called the N-clasp. The N-clasp extends from the thumb domain, causing DNA to be completely encircled near the primer terminus, likely stabilizing pol κ bound to not only matched, but also mismatched and lesion-containing termini (Fig. 2c) [154]. Mutational analysis has shown that the N-clasp facilitates extension from mismatched and likely damaged, bases [154]. It has been proposed that pol κ can extend a mismatch by locking the N-clasp around the primer-template, keeping it engaged on the distorted sugar phosphate backbones long enough for the misaligned 3ʹ OH to correctly re-position for nucleophilic attack [154].

Even though pol κ lacks proofreading activity, it has a moderate processivity and a low to moderate fidelity with error frequencies ranging for 10⁻² to 10⁻⁴ [145, 146, 155]. A constrained active site is thought to mediate this, as well as extension, by accommodating only one traditional Watson–Crick base pair at a time [153, 154]. While pol κ may be better at extending from lesions than bypassing them, biochemical and kinetic studies have shown that purified pol κ is capable of inserting nucleotides opposite several templating lesions, albeit with varying efficiency. As one example, studies show efficient and accurate bypass for N2-adducted guanines via dCTP insertion, with larger adducts being physically accommodated by a gap in the pol κ structure between the catalytic core and PAD domain (Fig. 2c) [156]. A recent crystal structure of a pol κ ternary complex, with a DNA substrate containing a bulky BP-dG adduct and incoming dCTP in the active site, provided insight into a domain arrangement that opens the active site at the minor groove to accommodate the damage. Furthermore, this structure supported the previously reported indispensable role of the N-clasp during lesion bypass (Fig. 2c) [157]. Another recently published ternary structure reveals a more ordered N-clasp domain allowing higher resolution insight into the function of this flexible domain [158]. The flexibility and stability of these domains uniquely suit pol κ for the specific translesion functions of bypass and extension.

DNA polymerase Rev1

In 1996, the yeast DNA polymerase Rev1 was detected to have dCMP transferase activity [159]. Since this protein had previously been shown to be required for damage-induced mutagenesis, it was proposed that this newly identified function was important for mutagenic TLS. However, Rev1 was not officially recognized as a DNA polymerase until 1999, when deoxynucleotidyl transferase activity was detected in several enzymes homologous to Rev1 [94]. Interestingly, Rev1 differs from other polymerases in its selectivity of nucleotide incorporation. While replicative polymerases incorporate the correct nucleotide opposite all four template bases with nearly equivalent efficiencies, Rev1 displays a strong preference for the incorporation of cytosine (C) opposite a templating guanine (G) [159].

The yeast Rev1 ternary complex structure (with DNA substrate and incoming nucleotide) provided the first structural basis for its highly specific mechanism, revealing a lack of hydrogen bonding between a templating G and an incoming dCTP in the Rev1 active site [160]. Rather, the templating G is removed from the active site, while the incoming dCTP pairs with an arginine side chain (Arg324) (Fig. 2d). Publication of a human Rev1 crystal structure revealed that the key elements of this mechanism are conserved in the human enzyme [161]. In the human Rev1 structure, Leu358 is inserted into the DNA helix, displacing the templating G, which consequently swings out of the helix and into a hydrophobic pocket (His774, Lys681, and Phe525). This extrahelical position is stabilized via two hydrogen-bonding interactions between its Hoogsteen edge and main-chain amides of a protein loop (Fig. 2d, green). As in yeast, the Watson–Crick edge of the incoming dCTP pairs with protein residue Arg357, making this a protein-template-directed mechanism. The pattern of hydrogen bonding for both the templating G and incoming dCTP is such that it favors the identity of these two specific bases. In both positions, the substitution of another base would result in loss of hydrogen bonding, as well as electrostatic or steric repulsion [160]. Therefore, these highly specific hydrogen-bonding interactions reveal the mechanism of Rev1 specificity for dCTP incorporation opposite a templating G. Rev1 uses the same protein-template-directed mechanism of DNA synthesis to insert dCTP opposite damaged guanines. Recently, structures of yeast Rev1 in complex with substrates containing DNA lesions, such as BP-N²-dG [162], abasic sites [163], and γ-HOPdG adducts [164], have allowed for the visualization of this mechanism in the context of damage (Fig. 2d, white). The stabilization of the templating damaged G via Hoogsteen hydrogen bonding ensures its rotation into the extrahelical position where the bulky adduct can be accommodated in the major groove, removed from the active site. Therefore, hydrogen bonding of the templating G ensures accommodation of the adduct, while hydrogen bonding of the incoming dCTP ensures correct insertion [162]. Thus, these structural studies provide a basis for Rev1′s unique ability to dictate both the incoming and templating bases to ensure error-free bypass of damaged Gs.

While the mechanism is conserved, the yeast and human Rev1 structures differ in that the human Rev1 catalytic domain contains two large inserts in the palm and fingers domains, making it significantly larger [161]. The 40-residue palm domain insert, which extends away from the active site and is largely disordered, has been proposed to serve as an additional site for TLS co-factor binding and/or protein–protein interactions. Rev1 has been shown to have an important non-catalytic scaffolding role in TLS, simultaneously interacting with several polymerases and potentially facilitating polymerase switching at lesions and multi-polymerase TLS [165, 166]. The 54-residue fingers domain insert supports the selectivity of Rev1 for dCTP insertion, contributing to the hydrophobic pocket by acting as a “flap” to stabilize bulky N2-dG adducts in the template position. As both inserts are partially unstructured, it has been proposed they become ordered upon additional interactions with protein or bulky DNA lesions, respectively [161]. Overall, fitting with the theme of this review, these structural minutia collectively provide Rev1 with an incredibly unique TLS ability which is vital to cellular function.

X-family

The X-family is comprised of four small monomeric DNA polymerases involved in DNA repair: polymerase beta (pol β), polymerase lambda (pol λ), polymerase mu (pol μ), and terminal deoxynucleotidyl transferase (TdT). Pol σ (formerly pol κ) is considered a distant relative of the X-family; however, we have chosen not to include a discussion of pol σ due to the absence of significant structural characterization [167, 168]. Each X-family DNA polymerase plays a unique role in DNA repair. Pol β primarily functions as a gap-filling polymerase during the base excision repair (BER) of oxidative base damage [15]. The remaining X-family polymerases contribute specialized functions within the double-strand break (DSB) repair pathways of non-homologous end-joining (NHEJ) and V(D)J recombination [4, 169]. To provide specificity for this wide range of DNA substrates and biological functions, all X-family polymerases utilize an “8 kDa domain” (8kD, Fig. 3a, blue) to bind to the downstream ends of gapped DNA substrates. Additionally, pol λ, pol μ, and TdT contain an N-terminal “BRCA1 C-terminus” (BRCT, Fig. 3a, magenta) domain that facilitates their recruitment to the NHEJ and V(D)J pathways for DSB repair. Below, we further discuss these domains as well as other structural details which allow the X-family members to perform their distinct functional roles.

Fig. 3.

The different structural features of the polymerase X-family members and their substrates. a A representative pre-catalytic structure of each X-family member highlighting the similarities and differences between the DNA substrate (orange), the incoming dNTP (yellow sticks), and the structural domains: Catalytic (gray), BRCT (magenta), 8 kD (blue), and Loop 1 (green) (left to right) 2FMS, 2PFP, 5TXX, 4I27 b Structural comparison of Pol μ (gray), Loop 1 (green) and DNA (orange) between the apoprotein (4LZD) and DNA-bound (5TXX) structures. c Structural comparison of Loop 1 (green) in TdT (gray) with different DNA (orange) substrates (left to right) 4I27, 4QZ8 and 5D46

DNA polymerase beta (pol β)

In 1971, Chang and Bollum discovered the presence of a small molecular weight DNA polymerase in mammalian cells [170]. At only 39 kDa, this polymerase, named DNA polymerase beta (pol β), is still considered one of the simplest cellular DNA polymerases. Pol β, which exhibits moderate fidelity (Table 1), shares catalytic features with the high-fidelity replicative polymerases; for instance, both families’ exhibit rate-limiting product release and a conformational change of the catalytic domain upon nucleotide binding [171–173]. Biologically, pol β plays dual roles in the nuclear BER pathway, where it both cleaves out the 5′ dRP left by AP endonuclease cleaveage and fills the resulting gap [174]. Additionally, pol β has also been implicated in the mitochondrial BER and alternative-NHEJ pathways [175–177]. The biological implications and precise role of pol β in these scenarios remain to be fully elucidated.

Pol β is comprised of two domains, an N-terminal 8 kDa domain and a C-terminal 31 kDa catalytic domain, both of which contain key structural elements mediating its functional roles during BER (Fig. 3a) [11]. During BER, pol β binds its substrate via two helix-hairpin-helix (HhH) motifs, one a part of the 8 kDa domain and the other a part of the catalytic domain [15]. This places the lyase active site of the 8 kDa domain in position to excises the 5ʹ dRP [178–180]. The HhH motifs also stabilize a 90° bend of the DNA, giving the polymerase access to the 3′-primer terminus, and allowing the incoming nucleotide entry into the nascent base pair binding site [15]. The resultant pol β:DNA binary complex is in an “open” conformation. Selection of a correctly matched nucleotide occurs via an induced fit mechanism, requiring a transition to a “closed” conformation in which the N-helix (fingers) realigns the active site residues and metal ions with the DNA duplex and incoming nucleotide [172]. Once the active site is arranged into its “closed” catalytically competent conformation, the nucleotidyl transferase reaction is carried out, where ultimately the O3ʹ of the primer terminus performs a nucleophilic attack on the alpha-phosphate via catalytic metal coordination and proton abstraction by Asp256 [181–183].

Equipped with the structural simplicity of only two domains and a low molecular weight, pol β is an ideal candidate for X-ray crystallography. Consequentially, pol β was the first member of the X-family (and first eukaryotic polymerase overall) to be examined at atomic resolution, and has continued to serve as a model DNA polymerase for structural characterization of the near-universal nucleotidyl transferase mechanism used by DNA polymerases for DNA synthesis [11, 17, 181, 184, 185]. With the development of time-lapse X-ray crystallography with pol β, the entire catalytic cycle of the nucleotidyl transferase mechanism has been observed [17, 186]. Importantly, this system has been exploited to reveal how DNA polymerases process non-canonical DNA substrates such as mismatches, damaged nucleotides, anti-viral drugs, and therapeutic nucleotides [17, 86, 181, 185, 187–189]. This comprehensive use of pol β to study the polymerase mechanism has resulted in a large body of literature, including 339 X-ray structures deposited in the PDB, focused on pol β. With this in mind, we have kept this section brief and direct readers to Ref. [15] for a review dedicated to pol β.

DNA polymerase lambda (pol λ)

DNA polymerase lambda (pol λ) was first reported as a novel eukaryotic DNA polymerase in 2000 [190]. At the time, it was thought to play a role in meiosis; today, however, pol λ is primarily considered a gap filling polymerase and is associated with several DNA repair pathways [191–193]. Of the eukaryotic X-family members, pol λ is considered the most closely related to their common ancestor [194]. Consistent with this notion, pol λ functionally overlaps with both of its template-dependent X-family siblings (pol β and pol μ). Moreover, like pol β, pol λ participates in BER [192, 195, 196], and like pol μ, it participates in DSB repair though the NHEJ pathway [193] and has been implicated in TLS [191]. As the largest member of the X-family, pol λ has an 8 kDa domain, a catalytic domain, a Ser-Pro rich region, and a BRCT domain (Fig. 3a) [197, 198]. The 8 kDa domain contains its 5ʹ-dRP activity utilized during BER [195, 199], the Ser-Pro rich region serves as a target for post-translation modifications [200], and the BRCT domain is responsible for the recruitment of pol λ to NHEJ complexes [201]. Several catalytic features of pol λ activity, including its method of nucleotide discrimination and propensity to produce deletions, have been revealed by structural analysis and are discussed below.

The comparison of several X-ray crystallography structures representing different stages in the pol λ catalytic cycle indicate that the enzyme does not undergo a large-scale open to closed transition upon binding of the incoming nucleotide to the binary complex (pol λ:DNA), as has been observed with many other DNA polymerases [202]. In contrast, a more localized transition occurs involving a relocation of a key loop motif termed “loop 1″ and a corresponding shift of the template strand further into the active site to facilitate base pairing between it and the templating base. In addition, several active site residues reposition to form a binding pocket for the nascent base pair and establish interactions with the DNA minor groove that are important for base selectivity and catalysis. Consistent with the structural predictions, deletion of loop 1 results in significantly decreased fidelity, indicating that loop 1 plays a role in nucleotide selectivity [203].

Both NHEJ and long patch BER involve filling gaps longer than one nucleotide. Structural analysis of the pol λ in complex with a two-nucleotide gapped DNA substrate suggests that DNA binding by pol λ is directed largely by the 8 kDa domain, which anchors the polymerase at the 5′-end regardless of the conformation of the 3′-end [202]. Upon binding of the incoming dNTP, the 3′-primer-terminal nucleotide, as well as amino acid residues that form the active site and the nascent base pair binding pocket, assume an identical position as observed in the structure of the ternary complex with a one-nucleotide gap. The incoming dNTP binds opposite the 3′-template nucleotide of the gap, which is located in the active site. Though the gap is one nucleotide longer, the distance between the 3′-end of the gap and the 5′-end bound by the 8 kDa domain is the same as in the one-nucleotide gap structure. This is possible because the template strand is scrunched, placing the 5′-template nucleotide of the gap in an extra-helical conformation within an active site pocket created by three amino acid residues (Leu277, His511, and Arg514) [204]. Upon a translocation, pol λ subsequently inserts the second base in the 5ʹ templating position, filling the two-nucleotide gap. Of note, pol λ generates a high frequency of single-base deletions during DNA synthesis compared to other DNA polymerases [205]. The mechanism for producing single base deletions is thought to involve dNTP-induced misalignment of the template strand relative to the primer strand during catalytic cycling [206]. Therefore, the likelihood that pol λ will fill a two nucleotide gap with either one nucleotide (i.e., creating a single-base deletion) or two nucleotides, appears to be dependent on the stability of the misaligned intermediate [204].

DNA polymerase mu (pol µ)

DNA polymerase mu (pol μ) was first identified and biochemically characterized in 2000 [207]. This initial characterization showed that the enzyme has very high error rates on primer–template substrates and additionally possesses a template-independent activity, where it extends from a template lacking single-stranded DNA primer. Combined with the observation that it is expressed preferentially in secondary lymphoid tissues, this early characterization suggested that pol μ could act as a DNA mutator polymerase responsible for the somatic hypermutation of immunoglobulin genes. Today, in terms of metal utilization, template dependency, and indiscriminate use of both ribonucleotides and deoxyribonucleotides, pol μ is considered one of the most versatile DNA polymerases [208, 209]. As a result of this unique substrate specificity, pol μ specializes in the bridging of uncomplimentary DNA ends during NHEJ and V(D)J recombination, where this versatility allows for the impromptu selection of the most efficient nucleotide/metal ion combination.

Structural analysis has provided insight into the unique substrate specificity of pol µ, elucidating key features that facilitate its specialization in end-joining. The structure of pol μ is organized into three domains: a BRCT domain, an 8 kDa domain, and a catalytic domain. Relative to its X family siblings, pol β and pol λ, the pol μ 8 kDa domain contains fewer DNA contacts and lacks lyase activity [210]. However, there is an increase in the number of contacts between the upstream DNA and the helix-hairpin-helix (HhH) motif contained within the catalytic domain, which assists pol μ in binding non-complimentary DNA strands [210, 211]. Functionally, pol μ has very slow polymerization rates (0.006–0.076 s⁻¹) for the correct insertion of deoxynucleotides [212]. In fact, the rates of correct insertion for pol µ are in the range of those obtained for the misincorporation of nucleotides by other DNA polymerases. Additionally, pol μ exhibits an increased fidelity for ribonucleotides [209, 213]. Importantly, this provides an advantage during NHEJ, during cell cycle timepoints when deoxyribonucleotide concentrations are very low. To this point, recent structures reveal that pol μ binds and incorporates ribonucleotides with canonical active site geometry and minimal distortion of the DNA substrate or nucleotide [213]. In addition, complementary incorporation kinetics utilizing wild type and rationally designed mutants of pol µ indicate that ribonucleotide accommodation involves synergistic interactions with multiple active site residues not found in polymerases with greater discrimination [213].

Structures representing various stages of the catalytic cycle indicate that pol µ is structurally rigid, where both the DNA and the enzyme lack large-scale conformational changes during dNTP binding and catalysis [214]. One exception is loop 1, which is oriented towards the DNA binding cleft of pol μ in the apoenzyme structure and is repositioned upon DNA binding to facilitate a catalytically competent position (Fig. 3b) [214]. Additionally, the inherent flexibility of this loop in pol μ appears to allow the accommodation of the different DNA substrates required for both its template-dependent and -independent activities. Supporting this hypothesis, deletion of the loop substantially reduces template-independent activity [215]. Biologically, this dual capacity could be advantageous to resolve microhomology-mediated NHEJ reactions. A ternary structure of Pol µ on a 2-nt gapped DNA substrate revealed that in contrast to pol λ, pol µ “skips” the first available template nucleotide, instead using the template base at the 5′ end of the gap to direct nucleotide binding and incorporation. This divergence from canonical 3′-end gap-filling provides additional insights into polymerase substrate choices during NHEJ [216]. Recently, time-lapse crystallography was employed to further examine the mechanism of pol µ during DSB repair [20]. These structures confirmed, using natural substrates, the lack of large protein or DNA conformational changes during catalysis, and identified several localized shifts of active site residues during metal binding and product dissociation.

TdT (terminal deoxynucleotidyl transferase)

Discovered in 1960, terminal deoxynucleotidyl transferase (TdT) is one of the first mammalian DNA polymerases to be identified, yet it remains one of the most poorly understood [169, 217]. TdT predominately incorporates nucleotides in a template-independent manner, using only single-stranded DNA as the nucleic acid substrate [218]. Despite its intriguing ability to create genomic material de novo, a specific biological role for TdT remained elusive for several decades. However, it is now known that TdT is responsible for the random addition of nucleotides to single-stranded DNA during V(D)J recombination [219, 220]. Importantly, this ability of TdT to intentionally create randomized genetic material during V(D)J recombination is vital for the mammalian immune system, as it promotes the immunoglobin and T cell antigen receptor diversity required for the neutralization of potential antigens [221–223].

Similar to pol μ, the structure of TdT is exceptionally rigid and consists of a BRCT domain, an 8 kDa domain, and a catalytic domain [224]. The first crystal structures of TdT, one with an oligonucleotide primer and another with an incoming ddATP-Co²⁺ complex, revealed steric clashes between the catalytic site and loop 1 (Fig. 3c). This long, “lariat-like” loop was proposed to explain the inability of TdT to accommodate duplex DNA [225]. In addition, these binary complexes revealed that the 8 kDa domain of TdT contacts the thumb domain to form a ring-like structure in which the single-stranded primer lies perpendicularly to the axis of the hole on the palm domain, and dNTPs diffuse through the hole into the enzyme’s active site. Later on, ternary complex structures of TdT bound to ssDNA with an incoming nucleotide were solved. Of note, there was no observed open to closed transition upon nucleotide binding, however, transient metal binding (occurring just before and leaving right after the chemistry step) in the active site was proposed to potentially drive efficient catalysis [226]. The combination of evidence from both structural and biochemical data deemed TdT incapable of interacting with duplex DNA substrates [224]. However, this widespread belief that TdT only possesses template-independent activity has recently been challenged.

TdT was recently shown to also have template-dependent activity in DSB repair, similar to pol μ, in the presence of higher concentrations dsDNA [227]. A series of crystal structures were recently solved with TdT bound to DSB substrates representing an annealed DNA synapsis containing one microhomology base pair and one nascent base pair (Fig. 3c). These structures revealed that the N-terminal 8 kDa domain and loop 1 cooperate to bridge the two DNA ends, providing a templating base in trans [228]. The structures from this and an additional study by the same group imply that TdT can direct DNA synapsis pre-assembly and alignment [227, 228], similar to pol µ [211]. Moreover, the structures indicate that TdT can halt its template-independent activity and switch into its template-dependent activity upon sensing the second DNA strand bridging the DSB. This is mediated through a reorganization of loop 1, which stabilizes the overlapping DSB strands (Fig. 3c). Consistent with this model, mutational analysis determined loop 1 to be essential for TdT to accommodate the opposing DNA strand during its template-dependent activity. Overall, this supports an intriguing model where TdT can single-handedly provide diversity via its template-independent activity and then accuracy via the template-dependent activity during V(D)J recombination. However, the precise biological role of these two activities attributed to TdT remains uncertain.

A-family

The members of the A-family of DNA polymerases share homology with E. coli DNA polymerase I, encoded by the polA gene. Mammalian cells contain at least three A-family DNA polymerases: gamma (pol γ), theta (pol θ), and nu (pol ν). DNA pol γ is responsible for the high fidelity replication and repair of mitochondrial DNA. Pols θ and ν, both low-fidelity nuclear enzymes, have been suggested to play roles in alternative nonhomologous end-joining and in the repair of interstrand DNA cross-links, respectively. Thus, the A-family polymerases have both similar and diverse biological functions. Structures of pol γ have shed new light on how the enzyme can uniquely adjust its processivity and fidelity to be efficient at both replication and repair [229]. Additionally, the crystal structures of pol ν and pol θ substrate complexes have elucidated unique polymerase structural diversity and identified several features of their respective fingers domains that confer their specialized functional attributes and provide insight into their poor DNA-synthesis fidelity [230–232].

DNA polymerase gamma (pol γ)

A novel RNA-dependent DNA polymerase activity in eukaryotic cells was reported in 1970 [233, 234]. By 1975, this polymerase was officially designated as DNA polymerase gamma (pol γ), although its cellular function as the main replicative and DNA repair polymerase in the midochondira was still undiscovered [235]. In 1977, pol γ was identified in mitochondria [236] and, not long after, evidence for its functional role was obtained by a study utilizing isolated brain synaptosomes [237]. Later, disruption of the yeast pol γ gene (MIP1) [238] and inhibition of mitochondrial DNA (mtDNA) replication by pol γ antibodies [239], provided further evidence for the mitochondrial role of pol γ. Accordingly, pol γ is critically important for mtDNA maintenance, and defects in pol γ-mediated mtDNA replication result in the loss of cellular respiration and ultimately induce mitochondrial genetic diseases, including mtDNA depletion disorders such as Alpers syndrome [240]. Additionally, human pol γ is known to be more susceptible than the nuclear DNA polymerases to inhibition by Nucleoside Reverse Transcriptase Inhibitors (NRTIs), which are designed to treat HIV; therefore, pol γ is likely responsible for a majority of the cellular toxicity caused by this key class of antiretroviral drugs [241, 242]. Of note, recent evidence has been mounting that pol γ is not acting alone in mitochondria, and this topic covered in Ref. [243].

The mtDNA replicated by pol γ is generally arranged in covalently closed circles and requires distinct molecular machinery from that used to replicate nuclear DNA. Two mechanisms of mtDNA synthesis have been proposed. The first is a conventional synchronous model, similar to that used to replicate nuclear DNA, where leading and lagging strand synthesis occur simultaneously [244]. The second mechanism is a strand-displacement (D-loop) model, where leading strand synthesis advances approximately two-thirds of the way around the circle before lagging strand synthesis is initiated [245–247]. Atomic level details, disclosed by a crystal structure of the human DNA pol γ heterotrimer, have helped elucidate the unique mechanism of mtDNA replication [229]. As was previously proposed based on a ~ 17 Å cryo-electron microscopy study [248], a single pol γA catalytic subunit interacts asymmetrically with a pol γB dimer (making contact primarily with only one pol γB monomer), as shown in Fig. 4a. Due to steric clashes, the authors were unable to model the enzyme as an A₂B₂ tetramer, which would potentially place two polymerases at a replication fork (a requirement for synchronous DNA synthesis). In agreement with the structural data, recent evidence, including ChIP-seq mapping of the mitochondrial ssDNA-binding protein (mtSSB) to the D-loop and the retention of primers at the two origins in RNaseH1 deficient cells, also points to the strand-displacement model as the more favored model of mtDNA replication [249].

Fig. 4.

A-Family polymerases. a Domain and subdomain structure of pol γ (3IKM). b Ternary structure of pol θ (green) bypassing an AP site, highlighting the DNA contacts at the primer terminus (4X0P). c Overlay of pol ν open (yellow; 4XVL) and closed (green; 4XVK) binary complexes

The crystal structure revealed that pol γA adopts the canonical polymerase ‘right-hand’ configuration and undergoes a partial conformational change upon nucleotide binding. Although the overall fold of pol γA confirms its classification as a member of the A-family of DNA polymerases, pol γA uniquely possesses a large spacer domain (~ 400 residues) between its exonuclease and polymerase domains. This spacer contains a globular IP (Intrinsic Processivity) subdomain that provides the intrinsic processivity of pol γA and an extended AID (Accessory Interacting Determinant) subdomain that forms an interface with pol γB, further increasing the processivity of the holoenzyme (Fig. 4a). As a result, pol γA alone can synthesize ≥ 100-nucleotides prior to dissociation [250, 251], and its processivity is further enhanced by the presence of pol γB, which simultaneously accelerates the polymerization rate and suppresses the exonuclease activity [252]. Structurally, pol γB resembles class II aminoacyl tRNA synthetases and differs significantly from other processivity factors, such as sliding clamps and thioredoxin [229]. Structures of the ternary complex revealed that the dimeric Pol γB rotates 22° in ternary complexes, increasing the interaction between pol γA and the distal pol γB monomer (Fig. 4a) [253]. Importantly, this pol γA-distal pol γB interaction was shown to allosterically regulate the polymerase and exonuclease activities in the holoenzyme.

MtDNA is damaged by both endogenous and exogenous sources. Moreover, the leakage of electrons from the electron transport chain promotes the generation of reactive oxygen species, which are the major source of oxidative damage to mtDNA, and can result in deleterious mutations. Mitochondria are able to repair oxidative lesions due to their robust base excision repair (BER) system [254]. However, the lack of other efficient repair systems, such as mismatch repair and nucleotide excision repair (NER), results in the persistence of mutagenic lesions in mtDNA. Because unrepaired lesions can promote mtDNA mutations or even block mtDNA replication, the ability of pol γ to bypass such lesions by TLS (for their subsequent repair) is vital to maintaining the genetic integrity of mtDNA. Translesion DNA synthesis by the human DNA pol γ has been studied for a number of DNA lesions, including 7,8-dihydro-8-oxo-2′-deoxyguanosine, benzopyrene adducts, UV photoadducts, acrolein adducts, and several others [255].

DNA polymerase theta (pol θ)

Early biochemical characterization demonstrated the ability of DNA polymerase theta (pol θ) to bypass DNA lesions, hinting at a role in TLS [256, 257]. This was confirmed when it was shown to be involved in bypass of the oxidative lesion thymine glycol in vivo [256, 258]. Pol θ exhibits a very low fidelity of synthesis on normal DNA, similar to other TLS polymerases [259]. In addition to TLS, pol θ has recently been shown to play a central role in the alternative DNA double-strand break repair process termed microhomology-mediated end-joining [260–264]. Microhomology-mediated end-joining is a mutagenic and error-prone alternative to homologous recombination (HR) or NHEJ that utilizes short (2–6 bp) microhomologies to join two strands of dsDNA [265]. Furthermore, the ability of pol θ to extend ssDNA [266] and its 5′ deoxyribosephosphate lyase activity [267] point to roles in other various DNA repair pathways [256, 258, 266–269]. Additionally, depletion of pol θ leads to hypersensitivity to radiation [263, 270, 271] and decreased survival of cells deficient in homology-directed repair [260]. Fitting with its role in DNA repair and TLS, overexpression of POLQ, the gene encoding pol θ, is a predictive marker for multiple human cancers [260, 270, 272]. Underscoring this, inhibition of pol θ is a compelling chemotherapeutic strategy in multiple cancers [273].

Recently, crystal structures of the human pol θ polymerase domain were reported demonstrating how pol θ grasps the primer to bypass DNA lesions or extend poorly annealed DNA termini to mediate end-joining [231]. The structures revealed pol θ to have a core DNA polymerase fold with fingers, palm, and thumb domains. However, the structure identified unique, unresolved loops that likely confer its specialized functions. Because pol θ can bypass simple DNA lesions, it is noteworthy that the structure of the ternary complex (with DNA substrate and incoming nucleotide) exhibits extensive backbone contacts with the primer strand (Fig. 4b). Several of these contacts are unique to pol θ, including an essential contact for lesion bypass between Arg2254 and the primer terminus phosphate [231]. In addition, the closed pol θ substrate complex with an incoming ddGTP indicates that Asp2376 of the fingers domain interacts with Arg2254 of the thumb domain. This primer-grip strategy, coupled with the ability of pol θ to dimerize, could enable the enzyme to direct DNA synthesis from mismatched primer termini during repair of DNA double-strand breaks.

Pol θ is the only polymerase known to include a helicase domain. The pol θ helicase domain (pol θ-HLD) is a member of the superfamily II (SF2) helicases, most closely related to the Hel308 family of helicases, which are involved in the ATP-dependent 3′–5′ unwinding of lagging strands on replication forks [274–277]. Biochemical characterization of pol θ-HLD has demonstrated DNA-dependent ATPase activity but has not detected any helicase activity, although it is possible that the helicase activity employs an untested substrate or requires an accessory factor [260, 278, 279]. Recently, crystal structures of apo and nucleotide-bound forms of pol θ-HLD were reported. In common with the recently derived structure of the pol θ polymerase domain [231], the structure displays unexpected features corresponding to insertions in its amino acid sequence that appear to be responsible for its unique activities. Specifically, large hydrophobic regions containing polar contacts appear to enable the specific tetrameric formation of POLQ-HLD both in solution and in crystallo; suggesting that, at minimum, the enzyme acts as a dimer [230]. In light of this structural revelation, it has been proposed that two pol θ molecules could work together during the MMEJ annealing step, with one molecule operating on each side of a DNA break, thus preparing the annealed substrate for further processing by the polymerase domain [230].

DNA polymerase nu (pol ν)

DNA polymerase nu (pol ν) is a newly discovered A-family polymerase identified through its homology to the Drosophila Mus308 gene, which is thought to function in the repair of DNA interstrand cross-links [280]. The precise biological role of pol ν remains unknown, however, it has been shown to be uniquely prone to misincorporating dTTP opposite a template G [281, 282]. In addition, pol ν is capable of accurate bypass of major groove DNA lesions including thymine glycol and major groove DNA-peptide and DNA–DNA cross-links [283].

Pol ν is homologous to high-fidelity replicative polymerases, yet is error prone [282]. Instead of a simple open to closed movement of the O-helix (fingers domain) upon binding of a correct incoming nucleotide, recent structures of pol ν binary complexes (with DNA substrate) indicate it has an atypical open state and that closure requires the fingers domain to assume multiple conformations to contact the nascent base pair [232], see Fig. 4c. Importantly, in this unique open state, the O and Ob helices exclude the template from the active site while allowing the dNTP to bind, which may provide an explanation for the low fidelity of pol ν. In the closed conformation, residues of the fingers domain contact the nascent base pair. Of note, Lys679 is positioned over the nascent base on the major-groove side (Fig. 4c), and alanine substitution for this lysine results in a ~ tenfold increase in fidelity. This lysine may contribute to the stability of mismatches by forming hydrogen bonds to the non-Watson–Crick base pairs. Further analysis of these results points to a multifaceted role for Lys679 in substrate discrimination that is expected to be dependent on DNA sequence context [284].

Conclusion

DNA polymerases fulfill many vital roles in the cell ranging from the accurate replication of the genome, both nuclear and mitochondrial, to numerous functions associated with the repair and bypass of DNA lesions. To effectively execute these diverse roles, there are at least 15 mammalian polymerases, each specialized for different functions. While numerous biochemical experiments have aided in the characterization of the unique properties of these enzymes, it has largely been structural studies that have elucidated their specific mechanisms at the molecular level. Through the revelation of these individual mechanisms, insight into how DNA polymerases collectively work together to utilize and accommodate such a wide range of substrates has also been revealed. This review is an overview of all the mammalian polymerases, providing a brief introduction into each and discussing how their structure dictates their function. We aimed to provide insight from some of the hallmark structures and studies that have been done, as well as point to other studies and reviews where more specific information can be found.

As more information is learned about the diversity in DNA polymerase structure, function, and mechanism, it can be leveraged against human disease more effectively. Already, many existing cancer therapeutics work by modulating the DNA polymerase-mediated DNA damage response [285, 286]. For example, the perturbation of DNA polymerase activity can promote therapeutic resistance (via lesion bypass/TLS) or be utilized during combinatorial therapies with the design of specific inhibitors based on polymerase structure and mechanism. It is likely that our knowledge of DNA polymerase structure and function will continue to advance with the discovery of novel enzymes with a polymerase-like activity, novel interactions of accessory proteins, or crosstalk between known polymerases and pathways. This is evident from the recently discovered PrimPol, which in just the last few years has been shown to have both primase and polymerase activities [287]. Not currently classified into any of the mammalian DNA polymerase families, PrimPol is under active investigation in the field today.