The RNA/DNA oligonucleotide captured in the crystals adopts an A-form conformation, as expected. The thumb domain engages in multiple interactions with the RNA primer, both via hydrophobic contacts and polar interactions (Perera et al., 2013). Experiments in solution have shown that the extension of the RNA primer by pol α is limited to 10–12 nucleotides, which amounts to one turn of a helix. This observation led the authors to suggest a mechanism for termination of primer synthesis by pol α in which loss of specific interactions between the thumb and the RNA would trigger the polymerase to disengage from the DNA/RNA oligonucleotide, and allow a hand off to a replicative polymerase.

Having crystallized the enzyme in three states (apo, binary, and ternary) allowed the authors to overlay all three structural models. Pol α is the only eukaryotic family B DNA polymerase for which all three states were captured in a crystal structure. The structural superposition revealed that, in addition to the well-documented movements of the fingers and thumb subdomains accompanying substrate binding and nucleotidyl transfer, the palm subdomain itself undergoes a structural rearrangement (Perera et al., 2013). The authors propose that the different conformations of the palm domain could facilitate translocation of pol α along and beyond the RNA/DNA duplex. As mentioned above, loss of contacts to the RNA strand is predicted to trigger release of primer, which then becomes available for extension by pol δ or ε.

The proofreading activity is abolished in pol α, due to mutations in all four carboxylates (Asp114/Glu116/Asp222/Asp327 in RB69 gp43 correspond to Ser542/Gln544/Tyr644/Asn757 in a structure-based alignment) (Table 2). Moreover, the β-hairpin motif found in most polymerases of the B family (residues 246–267 in RB69 gp43) is replaced by a helical region in pol α (residues 667–676; 681–693) (Hogg et al., 2007). The β hairpin is part of the exonuclease domain and has been shown in T4 and RB69 pols to participate in the partitioning of the DNA primer between the polymerase and the exonuclease active site (Reha-Krantz, 1988; Stocki et al., 1995; Hogg et al., 2007). In the absence of proofreading activity it is not surprising that this motif was not retained in pol α. Residues His 684 and Phe 685 of the helical region in pol α stack with a thymine and guanine base, respectively, at positions -3 and -2 in the unpaired 5′end of the template (Perera et al., 2013). Thus, in pol α the region corresponding to the β-hairpin motif adopts a different fold (helices vs. β strands) and a different function (stabilizing the unpaired region of the template strand rather than facilitating active site switching). Since pol α is devoid of proofreading activity the question arises as to whether the short oligonucleotides are corrected, and if so, by which DNA polymerase. It appears that proofreading of the primers synthesized by pol α is performed by pol δ (Pavlov et al., 2006a).

The palm domain contains three conserved carboxylates (Asp608, Asp762, and Asp764). The two catalytic aspartates, Asp608 and Asp764, contact two metal ions (Ca²⁺) in the polymerase active site separated by 3.7 Å. Intriguingly a third metal was observed coordinated by the γ phosphate of the incoming nucleotide and Glu802, with Glu800 in the vicinity. Mutating both glutamates to alanine yielded a polymerase variant with reduced incorporation efficiency for both correct and incorrect nucleotides (Swan et al., 2009). At these amino acid positions, pol α and pol ε also have carboxylate residues (pol δ Glu800/Glu802 correspond to pol α Asp1033/Asp1035, and pol ε Glu945/Asp947). Whether these carboxylates play similar roles in pol α and ε remains to be investigated.

Human pol δ is a high-fidelity polymerase, catalyzing the nucleotidyl transfer reaction with an error frequency of 1 per 22,000 (Schmitt et al., 2009). Proofreading boosts the fidelity of the polymerase by a factor of 10–100 (Mcculloch and Kunkel, 2008; Prindle et al., 2013). Pol δ harbors a polymerase and exonuclease active site, separated by about 45 Å (Swan et al., 2009). DNA polymerases with proofreading activity are able to sense misincorporated nucleotides by contacting the minor groove of base pairs beyond the insertion site. The protein interacts with universal hydrogen bond acceptors at the N3 and O2 positions of purines and pyrimidines, respectively (Seeman et al., 1976; Doublié et al., 1998; Franklin et al., 2001). These hydrogen bond contacts are preserved when the base pair adopts a Watson-Crick geometry and lost in the event of a mismatch. In RB69 gp43, the contacts extend to the first two base pairs beyond the nascent base pair (Franklin et al., 2001; Hogg et al., 2004, 2005). The contacts are much more extensive in pol δ, extending to five base pairs post-insertion (Swan et al., 2009), which could contribute to its high fidelity.

As mentioned above, the β-hairpin segment from the exonuclease domain plays a critical role in the partition of the DNA between polymerization and proofreading sites in T4 and RB69 pols (Stocki et al., 1995; Hogg et al., 2007). In RB69 gp43 the β-hairpin motif adopts different conformations, depending on whether the complex was obtained with undamaged DNA (Franklin et al., 2001; Zahn et al., 2007) or DNA containing a damage (Freisinger et al., 2004; Hogg et al., 2004). It was fully visualized contacting both the primer and template strands in a complex with thymine glycol (Aller et al., 2011). Similarly, the β hairpin in pol δ protrudes into the major groove of the DNA and acts as a wedge between double-stranded DNA and the single-stranded 5′end of the template strand, which is stabilized by two aromatic residues Phe441 and Tyr446 (Figure 1) (Swan et al., 2009). The position of the β hairpin is consistent with a role in active site switching.

Mutations involved in cancer are mostly found in the exonuclease domain of pol δ and ε, emphasizing the critical role of proofreading in lowering the incidence of mutations (Church et al., 2013; Henninger and Pursell, 2014). One mutation in human colorectal cancer cells localizes to the fingers domain, R689W. The analogous mutation in yeast (R696W) results in a mutator phenotype (Daee et al., 2010). A mutation in the vicinity of Arg696 in the highly conserved motif B of the fingers subdomain of yeast pol δ (A699Q) also results in a mutator phenotype. This region of the fingers is in close proximity to the NTD. Mutating Met540 of the NTD to alanine abolishes the mutator phenotype of A699Q, illustrating that interactions between the fingers and the NTD can affect the fidelity of the polymerase (Prindle et al., 2013). Similarly in T4 and RB69 pols the NPL core motif, which involves residues from the N-terminal and palm domains, is in contact with the fingers domain and was shown to stabilize polymerase-DNA complexes (Li et al., 2010).

Pol ε differs from pol δ in that it does not require the DNA sliding clamp PCNA for high processivity (Hogg and Johansson, 2012). The palm domain of pol ε is substantially larger (380 residues) than that of pol α or δ (175 and 203 residues, respectively). The recent pol ε crystal structures revealed that insertions in the palm domain collectively form a new domain consisting of three β strands and two helices (residues 533–555; 682–760) (Hogg et al., 2014; Jain et al., 2014a) (Figure 1; Table S1). Deleting residues 690–751 resulted in a variant with decreased polymerase activity. Moreover, mutating positively charged residues (His748, Arg749, and Lys751) located in the vicinity of the phosphate backbone affected the processivity of the enzyme (Hogg et al., 2014). The extra domain originating from the palm was thus named the processivity or P domain, after its function. The base of the P domain harbors a metal binding site (see below) (Hogg et al., 2014; Jain et al., 2014a,b).

Unexpectedly solution studies revealed that the catalytic subunit of yeast polymerase ε itself contains an [4Fe-4S] cluster within its polymerase fold (Jain et al., 2014b), in addition to the [4Fe-4S] cluster in the CTD (Figure 2; Table 2). The second [4Fe-4S] cluster within pol ε suggests that this polymerase may be more sensitive to oxidative stress (Jain et al., 2014b). The crystal structures of pol ε, however, did not reveal a [4Fe-4S] cluster in the polymerase domain (Hogg et al., 2014; Jain et al., 2014a; Zahn and Doublié, 2014). Two of the cysteines residues are disordered in the structural models and the resulting metal binding site appears to bind zinc (Hogg et al., 2014; Jain et al., 2014a). Substitution of a [4Fe-4S] by a non-native zinc in metal-binding proteins is not unusual (Netz et al., 2012) as [4Fe-4S] clusters are labile. Visualizing the [4Fe-4S] within the polymerase domain of pol ε may necessitate anaerobic conditions.

In any DNA polymerase harboring both polymerase and exonuclease activities the bound DNA is in equilibrium between the two active centers (Beechem et al., 1998). The concentration of incoming nucleotide and the presence of a damaged base or mispair are two factors that influence the transfer of DNA from the polymerase activate site to the proofreading active site. Polymerases monitor the minor groove side of the newly formed base pairs and interact with the universal H bond acceptors, O3, and N2, as a way of checking for mismatches (Seeman et al., 1976; Franklin et al., 2001). A unique feature of pol ε is the contact to the major groove side of the nascent base pair via a residue from the exonuclease domain, Tyr431. Further analysis is warranted to elucidate the potential role of this tyrosine in the high fidelity of pol ε.

In pol δ the β-hairpin segment inserts itself in the DNA and acts as a wedge between single-stranded and double-stranded DNA (Swan et al., 2009). In E. coli DNA pol II, the insertion of a β barrel shifts the position of the β hairpin in such a way that polymerization is favored over proofreading (Wang and Yang, 2009). This modification presumably allows this polymerase to carry out translesion synthesis extension. Since pol ε is an accurate DNA polymerase the assumption before knowledge of the crystal structure would be that the β hairpin should be closer to that of pol δ than that of E. coli Pol II. Surprisingly, the β-hairpin motif in pol ε is truncated, too short to contact the DNA (Figure 1). Which protein motif, then, might be facilitating active site switching upon sensing of a mispair? The P domain is a good candidate, because of its contacts to both primer and template strands; residues from the P domain could sense replication errors and thus may help facilitate active site switching.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.