Activity and fidelity of human DNA polymerase α depend on primer structure

DNA polymerase α (Polα) plays an important role in genome replication. In a complex with primase, Polα synthesizes chimeric RNA–DNA primers necessary for replication of both chromosomal DNA strands. During RNA primer extension with deoxyribonucleotides, Polα needs to use double-stranded helical substrates having different structures. Here, we provide a detailed structure–function analysis of human Polα's interaction with dNTPs and DNA templates primed with RNA, chimeric RNA–DNA, or DNA. We report the crystal structures of two ternary complexes of the Polα catalytic domain containing dCTP, a DNA template, and either a DNA or an RNA primer. Unexpectedly, in the ternary complex with a DNA:DNA duplex and dCTP, the “fingers” subdomain of Polα is in the open conformation. Polα induces conformational changes in the DNA and hybrid duplexes to produce the universal double helix form. Pre-steady-state kinetic studies indicated for both duplex types that chemical catalysis rather than product release is the rate-limiting step. Moreover, human Polα extended DNA primers with higher efficiency but lower processivity than it did with RNA and chimeric primers. Polα has a substantial propensity to make errors during DNA synthesis, and we observed that its fidelity depends on the type of sugar at the primer 3′-end. A detailed structural comparison of Polα with other replicative DNA polymerases disclosed common features and some differences, which may reflect the specialization of each polymerase in genome replication.

Three DNA polymerases are required for genome replication in eukaryotes: DNA polymerase ␣ (Pol␣), 5 Pol⑀, and Pol␦ (1). All of them belong to the B-family and fulfill the different tasks. Pol⑀ synthesizes most of the leading strand, Pol␦ is mainly involved in synthesis of the lagging strand, and Pol␣ generates the primers for both polymerases (2,3). Pol␣ alone cannot synthesize DNA primers de novo and relies on RNA primers created by primase. Pol␣ works in a tight complex with primase, called the primosome (4,5). Synthesis of the chimeric RNA-DNA primers by the primosome is highly coordinated by autoregulation through the alternating activation/inhibition of two catalytic centers, which is mediated by the C-terminal domain of the primase accessory subunit (6). Relatively low fidelity of Pol␣, which does not possess a proofreading activity, results in mutational hot spots predominantly on the lagging strand (7). In addition to the established role of primosome in nuclear replication, it is involved in formation of hybrid DNA:RNA duplexes in the cytosol, which are important for regulation of the type I interferon response (8). Pol␣ is a direct target of an anti-tumor toxin CD437, an attractive anti-cancer lead molecule, which induces apoptosis selectively in cancer cells (9).
Human Pol␣ (hPol␣) is composed of two polypeptides: the catalytic subunit (p180) and the accessory B-subunit (p70), with calculated molecular masses of 166 and 66 kDa, respectively. p180 contains two domains, the catalytic (residues 338 -1250) and the C-terminal (Pol␣ CTD, residues 1266 -1462) domains, which are flexibly connected by a 15-residue-long linker (4). The catalytic domain possesses DNA-polymerizing activity but has no proofreading exonuclease activity, in contrast to other replicative DNA Pols, ␦ and ⑀ (10). Pol␣ CTD connects the catalytic domain with p70 and primase and contains two conserved zinc-binding modules, where each zinc ion is coordinated by four cysteines (6,11,12). The N terminus of p180 (residues 1-337) is predicted to be poorly folded and does not participate in primer synthesis. The structural information for this region is limited to a small peptide in the catalytic subunit of yeast Pol␣ (yPol␣; residues 140 -147) that interacts with the replisome (13).
In this work, we use the structural and kinetic approaches to analyze hPol␣ interaction with the template:primer and dNTP and the effect of the primer structure on hPol␣ catalysis, processivity, and fidelity.

Overall structure of the catalytic domain of human DNA polymerase ␣ in complex with a DNA template, RNA, or DNA primer, incoming dCTP, and divalent metal ions
The structures of the ternary complexes of p180core (the part of hPol␣ containing the catalytic domain, residues 335-1257) with dCTP and DNA:RNA or DNA:DNA duplexes have been determined at 2.2 and 2.95 Å resolution, respectively ( Table 1). The p180core adopts the universal "right-hand" DNA polymerase fold consisting of five subdomains (14), which encircle the active site (upper schematic in Fig. 1A): N-terminal (residues 338 -534 and 761-808), exonuclease (Exo; residues 535-760; inactive), palm (residues 834 -908 and 968 -1076), fingers (residues 909 -967), and thumb (residues 1077-1250) domains. The palm is the most conserved subdomain in DNA polymerases and responsible for catalytically proficient positioning in the active site of two substrates to be polymerized: a template:primer and incoming dNTP. The palm harbors two catalytic aspartates (Asp 860 and Asp 1004 ), which coordinate two divalent metal ions important for dNTP binding and nucleophilic attack on its ␣-phosphate by the primer's 3Ј-OH (15). The thumb makes additional contacts with a template:primer, and their pattern significantly differs between B-family Pols (see below). The L-shaped thumb has two parts: the all-helical base adjacent to the palm and the tip with mixed ␣/␤-fold (residues 1135-1207), which packs against the Exo. Pol␣ lost proofreading exonuclease activity by substitution of the catalytic res-idues in the Exo subdomain during evolution (10). The fingers are composed of two antiparallel ␣-helices (Fig. 1A), which make additional contacts with the triphosphate moiety of dNTP and oscillate between the open and closed conformations. The thumb encloses the circle of p180core subdomains around the active site by connecting the palm with Exo (Fig.  1A). The interactions at the interfaces of the thumb with the palm and Exo have a lower amount of hydrophobic contacts in Pol␣ compared with Pol␦ and Pol⑀, which allows the thumb to rotate relative to the palm. Such rotation of the thumb away from Exo was observed in apo-forms of the human primosome and yPol␣ (6,16).
The alignment of hPol␣ ternary complexes containing DNA: DNA and DNA:RNA duplexes with a root mean square deviation (RMSD) of 0.86 Å for 648 C␣ atoms shows good superposition for the first four bp from the primer 3Ј-end and for p180core subdomains except the thumb and fingers (Fig. 1B). It is likely that the crystal packing affected the conformation of these subdomains in the hPol␣/DNA:DNA/dCTP complex. The thumb interacts with the palm and the N-terminal subdomain of one neighboring molecule and with Exo of the other neighboring molecule. The fingers are stabilized in the open conformation due to interaction with the N-terminal and Exo subdomains of the neighboring molecule. Significant difference in the position of the template:primer for bp 5-11 from the primer 3Ј-end (the bottom part of Fig. 1B) is due to steric hindrance with the thumb tip, which moves closer to the template upon the outward rotation of the thumb. In addition, the blunt end of the DNA:DNA duplex is packed against the N-terminal domain of the other molecule. In support of the idea that crystal packing affected the conformation of the most flexible subdomains in the structure of p180core/DNA:DNA/dCTP, the  Fig. 1B). Accordingly, hPol␣ has similar affinity for DNA and hybrid duplexes (17).
In the ternary complex containing DNA:DNA and dCTP, the fingers have the open conformation as in the binary complex hPol␣/DNA:DNA ( Fig. 2A) and in the complex hPol␣/DNA: RNA/aphidicolin (Fig. 2B). In all structures of yeast and human Pol␣, only two conformations of fingers were observed: open and closed (16 -18). This indicates that intermediate positions of fingers are energetically much less favorable. Upon closing, one of the fingers ␣-helices slides along the N-terminal subdomain toward to Exo (Fig. 2C). Analysis of intersubdomain interfaces in the open and closed forms revealed the switching of Ile 944 of the fingers between two adjacent pockets. One pocket is formed by Asn 760 , Pro 763 , Leu 764 , and Gln 767 of the N-terminal subdomain. The other is formed by Leu 764 of the N-terminal subdomain and by Leu 700 , Ile 701 , and Leu 759 of Exo. The open and closed conformations are also stabilized by a hydrogen bond between the side chain of Gln 941 and the mainchain oxygens of Leu 570 and Cys 757 , respectively.
In the structures of both ternary complexes, all three disordered regions (residues 674 -677, 810 -833, and 883-895) are situated near the template:primer. The loop 878 -903 origi-nates from the same site of the palm as the processivity domain (P-domain), which is found in Pol⑀ and provides additional contacts with a template:primer (Fig. 3). The linker 808 -837 between the N-terminal and palm subdomains of p180core corresponds to the shorter linker in Pol⑀, which adds two ␤-strands to the stem of the P-domain. In Pol␦, the corresponding region is also shorter than in Pol␣ and well-structured in the crystal (shown in blue in Fig. 3).

Interaction of hPol␣ with dNTP
The dNTP-binding site of B-family Pols is located at the junction between the palm and the fingers (Fig. 4A). Upon dNTP binding, the latter subdomain is able to change its conformation from open to the closed one, which results in stabilization of the precatalytic ternary complex Pol/template: primer/dNTP. Despite the open conformation of fingers in the complex of hPol␣ with a DNA duplex, dCTP occupies the dNTP-binding site (Fig. 4A). To our knowledge, this is the first structure of the ternary complex of a B-family DNA polymerase with a DNA duplex and dNTP where the fingers are open. This structure catches the transient step of dNTP binding followed by fingers closing and phosphodiester bond formation. It revealed that fingers interact with dNTP even in the open conformation. Asn 954 , which establishes the water-mediated contact with the ␤-phosphate in the closed-fingers conformation, directly interacts with this phosphate in the open-fingers state ( Table 2). Interaction of Lys 950 with dCTP switched from the ␣- A, overall view of the p180core/DNA:RNA/dCTP complex. The carbons of dCTP, DNA template, and RNA primer are colored yellow, marine, and purple, respectively. B, alignment of p180core ternary complexes containing RNA or DNA primers (PDB codes 4QCL and 6AS7, respectively). p180core, template:primer, and dCTP are colored gray in the complex containing DNA:DNA. In the bottom panel, DNA:DNA from the binary complex (PDB code 5IUD), which is superimposed onto p180core/DNA:RNA/dCTP with RMSD of 1.08 Å for 762 C␣ atoms, is colored orange (for clarity, double helices are shown separately from p180core).  (Table 2). In yPol␦, Asn 705 makes an H-bond with the ␤-phosphate. In yPol⑀, both water-mediated contacts observed in the hPol␣ ternary com-plex are changed to direct interactions. Thereby, only two residues of Pol␣ fingers (Arg 922 and Lys 950 in hPol␣) directly interact with dNTP in the closed state, whereas yeast Pol␦ and -⑀ use three and four residues, respectively.

Interaction of human
In hPol␣, the sugar and the base of dNTP fit into the hydrophobic pocket formed by Leu 864 , Tyr 865 , Pro 866 , Tyr 957 , and Thr 1003 (Fig. 4B). The conserved Tyr 865 defines a steric gate for rNTPs due to steric hindrance with the ribose 2Ј-OH, which is facilitated by a C3Ј-endo pucker. Upon fingers closing, the base of dCTP is fixed between Asn 954 and the 3Ј-nucleotide of the primer in the position that is optimal for pairing with guanine from the DNA template. The triphosphate of dCTP is secured in the active site by 11 bonds (Fig. 4A and Table 2), including two H-bonds with the main-chain nitrogens of Ser 863 and Leu 864 , four H-bonds with the side chains of Arg 922 and Lys 950 , and five coordinate bonds with two divalent ions, which in turn are coordinated by Asp 860 , Asp 1004 , and the main-chain oxygen of Phe 861 . hPol␣ also makes two water-mediated contacts with the phosphates of dCTP, using the side chains of Lys 926 and Asn 954 . The presence of Zn 2ϩ at the metal-binding site A in the hPol␣/DNA:RNA/dCTP complex is probably due to the missing hydroxyl at the primer 3Ј-end, which disrupts the octahedral coordination of Mg 2ϩ . In contrast to Mg 2ϩ , Zn 2ϩ is flexible with respect to the number of ligands that it can adopt in its first coordination shell (19). Moreover, Zn 2ϩ forms a stronger complex with carboxylates and the triphosphate group of nucleotides than Mg 2ϩ (20,21). Accordingly, Mg 2ϩ was missing at the site A in the crystal structure of yPol⑀ ternary complex, where the primer 3Ј-end also contained dideoxycytidine (22). Alignment of the hPol␣/DNA:RNA/dCTP complex with the ternary complexes of hPol␣ and yPol␦ containing a DNA primer shows good superposition for incoming dCTP and the residues coordinating dNTP and the catalytic metals (Fig. 4). This indicates a high structural conservation of dNTP-binding pocket in replicative Pols and that the type of the primer does not affect it.
Mutation of the conserved fingers residues, Arg 922 and Lys 950 , which interact with dNTP, dramatically reduced hPol␣  Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes activity in the presence of Mg 2ϩ (Fig. 5). Consistent with that, mutation of Lys 944 in yPol␣ is lethal (23). Substitution of Asn 954 , which makes a water-mediated contact with the triphosphate in the closed conformation, resulted in a modest effect on DNA polymerase activity. Kinetic studies have demonstrated that changes of Lys 950 primarily affect affinity for the incoming dNTP and the processivity of DNA synthesis, whereas their effect on catalysis was much less (24). Interestingly, activity of p180core mutants R922Q and K950S was almost completely restored in the presence of manganese, whereas the activity of the WT protein was weakly affected (Fig.  5). These data are consistent with studies of metal dependence of the activity of entire hPol␣ and its mutants with different substitutions of Lys 950 (24). Manganese provides a tighter complex with dNTP versus magnesium (25), allowing rescue of the compromised interaction of dNTP with the fingers in these mutants. The restoration of DNA-polymerizing activity of the finger mutants in the presence of Mn 2ϩ explains the wellknown mutagenic effect of this metal, which reduces Pol␣ fidelity by 2 orders of magnitude (24,25). Briefly, Mn 2ϩ -mediated binding of dNTP becomes less dependent on finger closure, an important selectivity step preventing mispairs.

Interaction of human Pol␣ with DNA:DNA and RNA:DNA double helices
Like other B-family DNA polymerases, Pol␣ makes continuous contact with the minor groove of the double helix of a template:primer (16,22,26,27). The palm interacts with phosphates of P 1 , P 2 , and T 2 -T 4 (Fig. 6B), which provides the optimal positioning of the primer 3Ј-end for catalysis. The flexible thumb secures the polymerase grip on the double helix by interacting with phosphates of T 7 , T 9 , and P 3 -P 5 . Two thumb residues, Arg 1082 and Thr 1140 , make double contacts with primer phosphates, using both the side-and main-chain atoms. Each of the N-terminal and Exo subdomains make only one contact with a template:primer. Interestingly, the position of T Ϫ1 phosphate is not fixed in Pol␣ (Fig. 6, A-D), whereas Pol␦ and Pol⑀ make a number of H-bonds with this phosphate (Fig. 6, E and F). In both ternary complexes of hPol␣, the side chain of Arg 834 extends into the major groove and packs against the base of T 3 .  Fig. 1 and gray, respectively. hPol␣ Phe 861 and yPol␦ Phe 609 , as well as two water molecules coordinated by zinc are not shown for clarity. dCTP from Pol␦ complex is shown as lines and colored dark brown. Atoms of magnesium, calcium, and zinc are presented as spheres at reduced scale and colored green, light orange, and slate, respectively. A different conformation of Asp 608 in yPol␦ is probably due to the coordination of a third calcium ion. Accordingly, Asp 860 of hPol␣ is oriented similarly to Asp 640 in yPol⑀ where Mg 2ϩ occupies the site B (22). In both panels, DNA:DNA and DNA:RNA duplexes are not shown for clarity.

Table 2 Amino acids of replicative DNA Pols making hydrophilic contacts with incoming dNTP
The residues that coordinate the divalent metals in the catalytic site are not listed. The residues marked by an asterisk use a main-chain nitrogen to establish an H-bond with the triphosphate. The residues making a water-mediated contact with the triphosphate are italicized. Residues shown in parentheses do not interact with dNTP in the specified complex.

Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes
On the other side of the double helix, the side chains of conserved Arg 1081 and Lys 1053 protrude into the minor groove and together with Arg 834 serve as rails for template: primer translocation. The reduced number of hPol␣-DNA contacts in the hPol␣/ DNA:DNA/dCTP complex ( Fig. 6A) is probably due to the crystal packing, which affected the thumb position (see above). This explanation is supported by the crystal structure of the hPol␣/DNA:DNA binary complex showing the same network of DNA-protein interactions as in the hPol␣/DNA:RNA/dCTP ternary complex (Fig. 6, B and C), with the additional contact between the P 1 phosphate and Tyr 1055 . The binding constants for p180core complexes with both duplexes are comparable (K d ϭ 58 Ϯ 2 nM for DNA:DNA and K d ϭ 101 Ϯ 6 nM for DNA:RNA; Table 3); the former value is in good agreement with a K d of 44 nM for the hPol␣/DNA:DNA complex reported by Copeland et al. (25). The discrepancy with data reported by Coloma et al. (17) (Table 3) might be due to different methods and experimental settings. Regardless, in both cases, the DNA and hybrid duplexes were compared at the same conditions, and their binding characteristics are similar.
As was observed in yeast Pols ␣, ␦, and ⑀ (Fig. 6, D-F), the invariant arginines 1081 and 1082 of hPol␣ trap the sugarphosphate backbone of the primer near P 4 . The side chain of Arg 1081 extends into the minor groove and packs its guanidinium moiety against the sugar (Figs. 6 (A-C) and 7A). Such a position of Arg 1081 is fixed by two H-bonds with Asp 1083 . The side chain of Arg 1082 is located between the phosphates of P 3 and P 4 and can interact with any of them. Binding of a hybrid duplex results in potential steric hindrance between Arg 1081 and the 2Ј-OH of the P 4 ribose (Fig. 7A) and in a local change of RNA conformation (Fig. 7B), which includes the unusual synconformation of P 4 (torsion angle ϭ Ϫ70.3°), 2Ј-endo pucker of its ribose, and unstacking between the bases of P 4 and P 5 . This bending of the RNA primer was also observed in the yPol␣/DNA:RNA/dGTP complex (16). It is notable that the orientation of the P 3 phosphate in the RNA primer is more favorable for hydrogen bonding with Arg 1082 and Arg 702 , compared with the P 3 phosphate of the DNA primer. The same position of Arg 1081 analog in ternary complexes of yeast Pols ␦ and ⑀ indicates a common mechanism of RNA sensing in replicative DNA polymerases.
In three reported complexes of Pol␣ with a hybrid duplex (16,18), the conformation of the P 1 phosphate is different compared with the complexes of hPol␣, yPol␦, and yPol⑀ with a DNA duplex (Fig. 8). The conformation of the P 1 phosphate was not affected by the addition of two deoxyribonucleotides to the 3Ј-end of the RNA primer, as seen in the yPol␣ ternary complex (Fig. 8B), which indicates that it depends on primer bending due to the presence of ribose at P 4 ( Fig. 7). Thus, accommodation of RNA and chimeric primers is accompanied by dislocation of the P 1 phosphorus, which results in disruption of the H-bond between this phosphate and the conserved Tyr 1055 ( Fig. 6B; this H-bond is replaced by the weaker watermediated interaction in the case of a DNA duplex).

Pol␣ does not exhibit typical polymerase burst kinetics
A pre-steady-state burst experiment was performed to provide insight into the relative rates of DNA polymerization steps in the kinetic mechanism of Pol␣. It has been frequently observed that under burst conditions (slight excess of template: primer over enzyme), many polymerases exhibit biphasic kinetics (28 -30). This biphasic pattern indicates that in the overall mechanism, the release of products is rate-limiting and slow compared with steps governing chemical catalysis. p180core (3 M) was incubated with either a 15-nucleotide DNA (D 15 ) or RNA primer (R 15 ) annealed to a 25-nucleotide DNA template (T 25 ; 9 M) to allow for the formation of the binary complex (Fig. 9A). The preincubated solution was rapidly mixed with dATP (200 M) using an RQF-3 rapid chemical quench apparatus. When the concentrations of products were plotted versus reaction time, an apparent biphasic curve was not observed but rather a linear formation of product for the elongation of both the DNA (k ss ϭ 104.3 Ϯ 3.1 s Ϫ1 ) and RNA (k ss ϭ 31.7 Ϯ 1.7 s Ϫ1 ) primer substrates (Fig. 9, B and C). Therefore, it appears that the Pol␣ catalytic core does not exhibit typical biphasic kinetics observed in reactions with main replicative polymerases, Pol␦ and Pol⑀ (31,32). This kinetic behavior indicates that for Pol␣, the rate of product release is equal to or faster than the rate of chemistry. These data are consistent with structural information showing weaker stabilization of Pol␣ fingers in the closed state compared with Pols ␦ and ⑀ (Table 2), which should increase the rate of finger opening and dissociation of a pyrophosphate/template:primer from a polymerase after catalysis. Conversely, the increased rate of finger opening may slow down Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes the overall rate of chemistry because finger closure stabilizes the triphosphate moiety of dNTP and, therefore, should facilitate the formation of a new phosphodiester bond. These results establish a similarity between Pol␣ and Pol, a polymerase implicated in base excision repair, which does not exhibit significant subdomain movements in response to nucleotide binding (33,34).

Pol␣ incorporates the first and second nucleotides into a DNA primer more efficiently than into an RNA
Single-nucleotide incorporation experiments were done under single-turnover conditions (excess Pol␣ over template: primer) to quantify the maximal rate of catalysis, k pol , and the apparent binding constant for the incoming nucleotide, K d . p180core (3 M) was incubated with a template:primer (100 nM) and rapidly mixed with varying concentrations of dATP (1-300 M) under rapid chemical quench conditions (Fig. 9D). For the DNA primer substrate, the k pol is 33.8 Ϯ 3.7 s Ϫ1 , the K d is 9.2 Ϯ 3.4 M, and the incorporation efficiency (k pol /K d ) is 3.7 M Ϫ1 s Ϫ1 (Table 4). For the RNA primer substrate, the k pol , K d , and k pol /K d values are 48.0 Ϯ 2.7 s Ϫ1 , 62.2 Ϯ 10.4 M, and 0.8 M Ϫ1 s Ϫ1 , respectively. It appears that Pol␣ incorporates a Among residues making van der Waals interactions with a template:primer, only the arginines contacting the P 4 sugar are shown for clarity. Residues highlighted in italic type can make a hydrogen bond with a template:primer in their alternative conformation. Ribose, deoxyribose, and dideoxyribose rings are colored red, gray, and olive, respectively. The residues interacting with a template:primer are color-coded according to the subdomains they belong to (see Fig. 1 for color-coding). The conservative amino acids are underlined. The residues marked by an asterisk use a main-chain atom to establish H-bond with a sugar-phosphate backbone. The pound sign indicates that the residue makes double contact with a template:primer, using both the side-and main-chain atoms. The type of the double helix (A, B, or intermediate AB) was determined for each dinucleotide step using Web 3DNA software (Rutgers University) (69). This software requires the presence of two consecutive dinucleotide steps of the same type to confidently determine the helix type, so the third dinucleotide step of DNA:RNA was downgraded by the program to unclassified. We assume that in locally distorted double helices, the presence of a single isolated step may take place, which is in line with data on yPol␣ (16). In the p180core/DNA:RNA/aphidicolin complex (PDB code 4Q5V), the pattern of interactions between hPol␣ and the template:primer is the same as shown in B except for the loss of the H-bond between T Ϫ1 and Ser 785 . Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes nucleotide more efficiently to a DNA primer by approximately a factor of 4.5. Strikingly, this discrepancy is primarily due to differences in the K d for the incoming nucleotide. This difference might be explained by reduced stability of the 3Ј-end of an RNA primer due to lack of interaction between Tyr 1055 and the P 1 phosphate (Fig. 6, B and D). As a result, increased dynamics of the primer 3Ј-end may affect dNTP loading to the active site through steric hindrance. In addition, the position of the primer 3Ј-end and dNTP binding might be affected by the presence of ribose at P 1 (see "Discussion"). Experiments were conducted to examine whether subsequent incorporations to a DNA primer were more efficient as compared with an RNA primer after the initial dNTP incorporation event. During the second incorporation, a 3Ј-hydroxyl of deoxyribonucleotide attacks the ␣-phosphate of incoming dNTP in both the DNA and RNA primer extension assays. Double-incorporation assays were conducted to estimate the kinetic parameters of a subsequent incorporation after the first incorporation. Experiments were designed to observe a dATP followed by a dTTP incorporation under single-turnover conditions. Preincubated solutions of p180core (3 M) and template:primer (100 nM) were mixed with a solution containing both dATP and dTTP. Because the kinetic parameters of the second incorporation were desired, concentrations of dATP were held at saturating amounts (100 M for DNA primer and 300 M for RNA, based on single-incorporation experiments), and concentrations of dTTP were varied from 1 to 300 M (Fig. 10A).
The appearance of the second incorporation product is complex due to its dependence on the first product being made; product concentration cannot simply be fit to a single-exponential curve like the single-incorporation assays. Thus, the KinTek Global Explorer program (35) was utilized to fit the data and provide an estimate for the k pol and K d of the dTTP incorporation (Fig. S1). For the second incorporation to a DNA primer, the k pol is estimated to be 65.8 s Ϫ1 , and the K d is 5.  (Table 5). This 6-fold difference in incorporation efficiency for DNA versus RNA for sequential dNTPs is comparable with the single-incorporation results. Contrary to the single-incorporation experiments, a major difference in k pol values is observed in the double-incorporation experiments. We hypothesize that the ϳ3-fold difference in k pol values may stem from the primer bending at P 4 in the hybrid duplex. The altered helix structure possibly affects the rate of any step in subsequent incorporations after the initial incorporation, such as translocation or necessary protein conformational changes. The reason for similar k pol values during the first incorporation is the preforma-  A, conformation of P 1 -P 2 in the complexes of hPol␣ with DNA and chimeric primers. The p180core/DNA:RNA/dCTP and p180core/ DNA:DNA complexes (PDB codes 4QCL and 5IUD, respectively) were aligned with RMSD of 0.42 Å using the palm. In the former complex, the RNA primer may be considered chimeric due to the presence of dideoxycytidine at the 3Ј-end. The carbons are colored gray in the complex of p180core with DNA: DNA. B, conformation of P 1 -P 2 in the complexes of yPol␦ and yPol␣ with DNA and chimeric primers, respectively. The ternary complexes of yPol␣ and yPol␦ (PDB codes 4FYD and 3IAY, respectively) were aligned with RMSD of 0.67 Å using the palm. The carbons are colored gray in the complex of yPol␦ with DNA:DNA. Blue dashed lines depict the H-bonds.
Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes tion of the enzyme/template:primer complex before the incorporation reaction is initiated.

Comparison of Pol␣ processivity on DNA:DNA and DNA:RNA duplexes
A processivity experiment was employed to examine multiple rounds of dNTP incorporation and look at the full extension of the primers and determine whether the rates of each incorporation were different over time between DNA and RNA primers. This analysis observes the k pol for each incorporation event as well as k off , which characterizes the dissociation rate of the Pol␣/template:primer complex. The assay was conducted under single-turnover conditions and saturating concentrations (240 M) of all dNTPs (Fig. 10, B and C). The products were plotted against time and fit to a processive mechanism using KinTek Global Explorer (Fig. 10, D and E; for visual clarity, only the first five incorporations are shown). The k pol and k off values for the first seven incorporations are shown in Table  6. Although this experiment provides only estimates of these rates, it is clear that there is a higher average k pol for the second to fourth incorporations during DNA primer extension, with an average value of 26.8 s Ϫ1 compared with 12.0 s Ϫ1 for RNA. As was shown above, the first incorporation does not show the real k pol values due to preformation of Pol␣/template:primer complex; that is why it was omitted from averaging.
Interestingly, we observe an increase in the rate of incorporation after the fourth incorporation into the RNA primer, with an average k pol for the fifth to seventh incorporations of 24.4 s Ϫ1 , which is similar to the value obtained on a DNA primer (22.2 s Ϫ1 ). This is particularly interesting in light of our structural results pertaining to the kink at P 4 in the hybrid duplex structure. This observation can be explained in structural terms by the fact that after four incorporations, the rigid and conserved DNA-binding cleft of Pol␣ no longer makes contacts with the ribonucleotides in the hybrid duplex. This increase in k pol agrees with our previous experiments demonstrating that DNA primer elongation is more efficient. Furthermore, to confirm that the increase from 12.0 to 24.4 s Ϫ1 after the fourth incorporation is significant, we simulated how the time course of processive elongation of an RNA primer would change if the fourth incorporation rate was modified to match the rate observed in the DNA experiment (Fig. 10E, dotted lines). Changing 8.3 s Ϫ1 to 19.3 s Ϫ1 results in a drastically different product distribution. The most prominent effects are the decreased accumulation of R 18 and faster production of R 20 because the rate of R 19 formation is no longer significantly rate-limiting.

Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes
It was also observed that the average k off value for DNA primer elongation is 7.0 s Ϫ1 compared to 2.3 s Ϫ1 for RNA (for the second to fourth incorporations). The higher k off value for the DNA duplex may be due to the tendency for it to adopt the B-form, which is slightly narrower than the template:primerbinding cleft. In contrast, the hybrid duplex is squeezed between Gly 841 and Gly 1076 ( Fig. 6B; see "Discussion"), which may potentially slow the dissociation and association of the duplex. Indeed, the estimates for k on values (0.12 and 0.02 nM Ϫ1 s Ϫ1 for DNA and the hybrid duplex, as a first approximation) calculated by dividing the average k off for the first four incorporations by the K d show a higher rate of complex formation with the DNA duplex. It is noteworthy that the interaction of the thumb with the distant part of a template:primer (T 5 :P 5 -T 9 :P 9 ) can also contribute to slower dissociation of Pol␣ from the hybrid duplex because this DNA-binding region is optimal for recognition of the double helix of mixed AB-form (Fig. 6B). Accordingly, two dinucleotide steps in this region of the DNA duplex have the mixed AB-form (Fig. 6C).
The ratio k pol /k off provides an estimate of processivity and indicates that on average (for the second to fourth incorporations) hPol␣ incorporates approximately three or four and six consecutive nucleotides before complex dissociation in the case of DNA and RNA primer, respectively. Using a DNA-trap, it was shown that hPol␣ incorporates six consecutive nucleotides before dissociation from the DNA duplex (36). Slightly higher processivity observed by Copeland and Wang's assay may be due to the absence of the monovalent salt in reaction (0.1 M NaCl was used in our experiments). In extension of the DNA primer, yPol␦ has the same processivity as hPol␣, whereas yPol⑀ catalytic core containing an unique processivity domain incorporates 27 nucleotides in one binding event, in the absence of monovalent salt in reaction (22). Interestingly, k off values are comparable for hPol␣ and yPol⑀, and the difference in processivity is mainly due to the higher incorporation rate of the latter. Higher processivity of Pol␣ on the hybrid duplex can prevent the enzyme from premature dissociation.

Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes
Additionally, the processivity of a polymerase can be qualitatively assessed by inspection of separated products on a gel. Comparing the processivity gels from Pol␣ (Fig. 10, B and C) and T7 DNA polymerase (30), the striking difference is the presence of primers that have not been fully extended with dNTPs at later points in the experimental time course. Through our experiments, it is observed that there is a significant population of primers that have been only singly elongated (D 16 ) at later time points even when the final product is finally produced (D 25 ). In contrast, once T7 DNA polymerase incorporates a dNTP into a primer, it continues to incorporate nucleotides into the same primer. In the direction of increasing time, the pattern on the gel for the Pol␣ can be described as extending a ladder upward, whereas T7 can be described as shifting a ladder upward. Processivity experiments have also been done with Pol⑀ (31), and the pattern of processive polymerization is similar to Pol␣. Although Pol⑀ is considered more processive than Pol␣, the striking difference between T7 and the former polymerases highlights the highly variable kinetic characteristics of polymerases.

The sugar at the primer 3-end affects Pol␣ fidelity
The potential of Pol␣ to generate 12 possible mismatches during the extension of a DNA or RNA primer was compared. This study revealed that p180core has pronounced difference in nucleotide selectivity during the addition of the first dNMP to the RNA and DNA primers (Fig. 11). It is relatively accurate on RNA but easily generates various mismatches when extending the DNA primer. The most prominent of them are A:dA, T:dG, G:dT, G:dG, and C:dT (the left letter indicates a nucleotide of the template). Moreover, p180core efficiently extends some mismatches (Fig. 11B, lanes 6, 7, 11, 13, and 16), even with noncognate nucleotides (lanes 6, 7, and particularly lane 16 showing two consecutive A:dC mismatches and their extension). Interestingly, Pol␣ is not able to extend the most prominent mismatch, A:dA, whereas it easily extends C:dA and G:dA by generating the next consecutive mismatch, A:dA (Fig. 11B,  compare lanes 6 -8). The results are consistent with kinetic studies of human and yeast Pol␣, showing that they discriminate against noncognate dNTPs and the mismatched primer terminus by a factor of 10 3 to 10 4 and 500, respectively (25,37,38). The discrimination mainly depends on the difference in affinity for a correct versus incorrect incoming nucleotide, thus explaining reduced selectivity in the presence of 0.2 mM dNTP in our experiments (this concentration is ϳ20-fold higher than concentration of dNTPs in mitotic cells (39,40)).
Intriguingly, Pol␣ becomes less accurate after incorporation of just one dNMP to an RNA primer. For example, no product was detected during the first step of RNA primer extension with dATP opposite the template adenine (Fig. 11C, lane 8), whereas in the same time frame, a significant amount of A:dA mismatch was generated during the second step of RNA primer extension, following incorporation of the first dAMP opposite the template thymine (lane 9). More efficient mismatch generation during the second step is not due to formation of the weak T:dA bp because the same effect was observed using other dNTPs (Fig. 11C, comparison of lanes 2 and 4 (dGTP), 11 and 12 (dTTP), and 15 and 16 (dCTP); appearance of the second and/or the following bands above the product indicates mismatch formation). Dramatic accuracy reduction after the insertion of just one dNMP indicates that Pol␣ fidelity depends on the presence of ribose at P 1 but not on the primer bending at P 4 (Fig. 7).
Initial burst experiments, performed on a broad time scale to establish the relative rate of dNTP incorporation by the p180core, revealed that Pol␣ incorporates an incorrect dNTP at a notable rate (Fig. 12). This is illustrated at later time points by the appearance of a second band that corresponds to a dATP

Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes
incorporation across from a template adenine. This observation was unexpected due to the prevalence of misincorporation within a time frame relatively close to correct incorporation. Although previous studies (25,41) have indicated that Pol␣ is a low-fidelity polymerase, to the best of our knowledge, this is the first direct observation of significant Pol␣ misincorporation in a millisecond time scale. These results further emphasize the benefits of utilizing a pre-steady-state kinetic approach.

Discussion
Currently, the structure of Pol␣ catalytic domain is thoroughly studied. Eight crystal structures have been reported for yeast and human orthologues, which include the apo-form, binary, and ternary complexes, containing DNA or RNA primer, as well as the structures of human primosome and the complex of hPol␣ with the natural inhibitor aphidicolin (6, 16 -18). In addition, the crystal structures of the CTD-B-subunit complexes are known for yeast and human Pol␣ (11,12). With regard to the catalytic subunits of Pol␦ and Pol⑀, only the structures of the ternary complexes have been determined for their catalytic domains (22,26,42) as well as the structure of the Pol⑀ CTD-B-subunit complex (43).

The role of conformational dynamics in DNA synthesis by Pol␣
Accumulated structural information for the catalytic domain of Pol␣ indicates that the N-terminal and exonuclease subdomains are firmly associated, whereas positions of other subdomains relative to the N-terminal part of p180core and each other are quite flexible (this work; also, see Refs. 6 and 16). Such flexibility of the subdomains responsible for substrate binding and catalysis defines the conformational dynamics of the polymerase during DNA synthesis. For example, significant mobility of the thumb might be important for accommodation and translocation of duplexes containing lesions, mismatches, and different types of primers: RNA, DNA, and chimeric RNA-DNA. Accordingly, Pol␣ has notable translesion activity and the ability to extend the mismatched 3Ј terminus of a primer (25,44).
Relatively weak interactions of the mobile part of the fingers with the other subdomains allow them to switch quickly between the open and closed states. High dynamics of the fingers is important for rapid closing after dNTP binding and opening after catalysis to release the products. The binding of dNTP results in the shift of equilibrium to the closed conformation, which in hPol␣ is primarily due to formation of two H-bonds between Arg 922 and the ␥-phosphate ( Table 2). The dNTP becomes trapped after fingers closing, promoting effective catalysis. Crystal packing may have helped to stabilize the open conformation of the fingers in the hPol␣/DNA:DNA/ dCTP complex because, in free energy terms, the open state should be less favorable versus the closed one in the presence of bound dNTP. The other known example of a DNA polymerase crystallized in the open ternary complex is Pol␤, which belongs to the X-family (45). That structure was obtained after mutating the conserved Arg 283 important for stabilization of the closed state, which shifted the equilibrium between two conformations of the helix N to the open one. Whereas the helix N fulfills a similar task as the fingers in Pol␣, it does not interact with a triphosphate of an incoming nucleotide.
This study revealed that Pol␣ fingers sense the triphosphate moiety of dNTP even in the open state ( Fig. 4A and Table 2). Interaction of open fingers with dNTP might be important for stabilization of the ternary complex during the first step of its formation and could trigger fingers closing. Currently, there is a lack of information on how the fingers of yeast Pol␦ and Pol⑀ interact with dNTP in the open conformation. Furthermore, the structures of these Pols in the apo-form are not reported, which precludes the possibility to model potential interactions between the residues of open fingers and the triphosphate. The absence of a rate-limiting step for product release revealed from pre-steady-state burst experiments reflects fast dynamics of Pol␣ fingers caused by weak stabilization of the closed conformation. This structural peculiarity of Pol␣ results in higher probability of fingers opening, facilitating mismatch accommodation at the nascent bp-binding pocket (see below). This finding may provide a key insight into the mechanism of the generation of mutation hot spots in the genome (7). Interestingly, all amino acid changes in hPol␣ conferring resistance to CD437, a potential anticancer molecule recently discovered to target Pol␣, are located at or near the interface between the fingers and the N-terminal/Exo subdomains (9).

Comparison of template:primer binding by replicative DNA Pols
All invariant residues of replicative DNA polymerases contacting the template:primer interact with the first five bp of the duplex (Fig. 6). The most interesting feature of these polymerases is the interaction of T 3 and P 2 phosphates with two conservative glycines of the palm (Gly 841 and Gly 1076 in hPol␣). These glycines determine the width of the double helix between the second and third bp from the primer 3Ј-end, with similar distances between their main-chain nitrogens for hPol␣ in complex with DNA:RNA (20.2 Å) and yeast Pol␦ and -⑀ in complex with DNA:DNA (20.4 and 20.3 Å, respectively). The other common feature of replicative polymerases is the absence of interaction with a T 1 phosphate (Fig. 6), which might be important for adjusting the position of the templating nucleotide to obtain optimal pairing with an incoming dNTP. In contrast to Pol⑀, the thumbs of Pol␣ and Pol␦ are engaged almost exclusively with the primer and do not interact with T 5 and T 6 . The  (22)). The low footprint of Pol␦ on the template is compensated by the double contact with the P 6 phosphate (Fig. 6E). In contrast, Pol␣ and Pol⑀ sense only five continuous nucleotides of the primer. The recently discovered P-domain of Pol⑀ allows it to make three additional contacts with a DNA duplex (Fig. 6F), one with a primer (P 8 ) and two with a template (T 2 and T 3 ), and two extra contacts are possible (22). These additional contacts and the extensive network of polar contacts between the thumb and the template explain the higher processivity of Pol⑀ catalytic domain compared with Pol␦ and Pol␣. In both ternary complexes of hPol␣, the side chain of Arg 834 located on the flexible loop extends into the major groove and packs against the base of T 3 . For comparison, yeast Pol␣ and Pol⑀ have no contacts with the major groove, whereas the ␤-hairpin of yPol␦ is anchored in it (26).
In solution, the RNA, DNA, and hybrid duplexes (DNA: RNA) prefer to be in the A-, B-, and mixed AB-form, respectively (46 -48). The A-form is wider and shorter than the B-form, with a diameter of 23 Å versus 20 Å and a rise per base of 2.6 Å versus 3.4 Å, respectively. DNA-protein interactions can affect the shape of the double helix, locally or globally (49). The conserved character of interaction with the first four bp determines the universal form of the double helix for this region in the ternary complexes of Pol␣, Pol␦, and Pol⑀, independent of the primer nature (Fig. 6). This is also true for the chimeric RNA-DNA primer (Fig. 6D), which reflects the initial steps of RNA primer extension by Pol␣. Replicative DNA polymerases transform the first two dinucleotide steps of DNA:DNA into the mixed AB-form and the third dinucleotide step of DNA: RNA into the B-like form (the latter is currently shown only for ternary complexes of Pol␣). Interestingly, in all ternary complexes of eukaryotic replicative DNA polymerases, the sugars of T 0 , T 1 , P 1 , and incoming dNTP have a 3Ј-endo pucker (or closely C2Ј-exo and C4Ј-exo puckers), which might be important for catalysis.
The alignment of hPol␣/DNA:RNA/dCTP and yPol␦/DNA: DNA/dCTP complexes (RMSD of 2.08 Å for 524 C␣ atoms) shows good superposition for the first four bp of the double helix, which is consistent with the same duplex form in this region as well as with high structural similarity for the palm and the N-terminal part of the thumb (Fig. 13). In contrast to Pol␣ and Pol⑀, Pol␦ has a short loop in the tip of the thumb (residues 895 PNYTNP 900 ; highlighted in blue), which allows it to interact with the P 6 phosphate using the main-chain nitrogens of Thr 898 and Asn 899 (Fig. 13). This loop also contains the conserved Tyr 897 , which interacts with P 5 and, together with Thr 898 and Asn 899 , would clash with P 6 and P 7 upon DNA:RNA binding by Pol␦. The additional steric gate residues for the wider AB-form of the hybrid duplex in yPol␦ are Leu 935 and Tyr 936 , which are located on the 3 10 -helix of the thumb tip and correspond to Ala 1188 and Ser 1189 of hPol␣, respectively (Fig. 13). This region may have a minor role in hybrid duplex sensing by Pol␦ due to Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes potential flexibility of the thumb. Consistent with this structural analysis, Pol␦ extends the DNA primer with significantly higher efficiency versus RNA (50). Similar to Pol␣, Pol⑀ does not interact with P 6 (Fig. 6, B and F) and efficiently extends primers containing up to five consecutive ribonucleotides at the 3Ј-end (51). Pol␦ also differs from Pol␣ and Pol⑀ (22) by the presence of a ␤-hairpin near the unprimed region of the template (Fig. 13).
In summary, all replicative DNA Pols have the same rigid template:primer-binding cleft for the first four bp (from T 1 :P 1 to T 4 :P 4 ), which is mainly formed by the palm residues (Fig. 6). The DNA-binding site distant from the catalytic center (for the duplex region spanning 5-10 bp from the primer 3Ј-end) significantly differs in B-family Pols and probably reflects the role of each polymerase in DNA replication and repair. For example, the low efficiency of RNA primer extension by Pol␦ as well as increased processivity of the Pol⑀ catalytic domain are mainly due to polymerase-specific structural peculiarities of this part of the DNA-binding cleft.

Structural basis of Pol␣ fidelity and infidelity
Pol␣ demonstrates relatively high fidelity during RNA primer extension with the first dNTP, whereas the incorporation of the following nucleotides is less accurate (Fig. 11). Surprisingly, the bulky purine:purine mismatches are among the most prominent, which is probably due to the ability of Pol␣ to easily fit them into the nascent bp-binding pocket. Current structural information suggests that at least four factors may assist in formation of the purine:purine mismatches: fingers opening (Fig. 4A), syn-conformation of T 0 stabilized by Arg 784 (17,18), absence of the ␤-hairpin ( Fig. 13; in Pol␦, the T Ϫ1 sugar and the T 0 phosphate are secured between the ␤-hairpin and the N-terminal subdomain), and weak interaction of Pol␣ with the template 5Ј-overhang (Fig. 6B), which allows more conformational freedom for T 1 -T Ϫ1 during mismatch accommodation. The exact mechanism of Pol␣ low fidelity is not known due to the absence of structural information for the WT B-family DNA polymerases in the mismatch-containing ternary complexes.
Intriguingly, the structure of the p180core/DNA:RNA/ aphidicolin complex (18) provides a clue for how the most prominent A:dATP mispair fits the active site of Pol␣. Indeed, all four factors listed above play a role in the accommodation of the bulky aphidicolin in place of the incoming dCTP, where the local change of template conformation (Fig. 14A) resulted in disruption of interaction between T Ϫ1 and Ser 785 . The modeling shows that Pol␣ can easily accommodate all purine:purine mispairs when the T 0 purine is in syn-conformation (Fig. 14B). In addition, A:dATP can fit the nascent bp-binding pocket when the templating adenine is in the anti-conformation. Small steric hindrance between two purines in this case might be relieved by the additional minor adjustments in the active site. This model of A:dATP mispair accommodation is applicable to the other mismatches. Kinetic studies of yeast Pols ␦ and ⑀ are consistent with the role of the ␤-hairpin in the fixation of the template position at T 0 , which should significantly reduce the chance to generate the bulky purine:purine mismatches. Among replicative Pols of eukaryotes, only Pol␦ has the ␤-hair-pin and generates the purine:purine mismatches with relatively low efficiency (32). Accordingly, exonuclease-deficient yPol␦ and yPol⑀ discriminate against the G:dG mismatch by a factor of 6 ϫ 10 4 (32) and 5 ϫ 10 3 (52), respectively.
The apparent low fidelity of Pol␣ during incorporation of the second and following dNMPs allows us to compare the enzyme with other well-studied polymerases. HIV-1 reverse transcriptase has been established as a low-fidelity polymerase (53). Its activity is beneficial for the virus, allowing for rapid mutation and development of resistance against antivirals. Although poor discrimination between nucleotides may be desirable for viruses, this trait is not particularly advantageous for host polymerases, which are responsible for perfectly replicating the genome. The low accuracy of Pol␣ might be important for the evolution of regulatory elements in eukaryotic genomes (7).  (PDB codes 4QCL and 5Q4V, respectively) were aligned with RMSD of 0.27 Å using the palm. B, modeling of the A:dATP mispair in the hPol␣ active site. DNA carbons are colored slate and gray in the case of the ternary complexes with dCTP and aphidicolin, respectively. The carbons of modeled adenines at T 0 and in place of dCTP are colored green and cyan, respectively. C, the modeled 2Ј-OH has steric hindrance with Thr 1003 . The coordinates of the p180core/DNA:RNA/ dCTP complex (PDB code 4QCL) were used for modeling (the sugar of the 3Ј-dideoxycytidine has a 3Ј-endo pucker).

Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes
Pol␣ has a significant contribution to the mutation rate in vivo, despite the fact that only 1.5% of Pol␣-synthesized DNA has been shown to be retained in the mature genome (7). Poor fidelity in host polymerases is not limited to Pol␣ and translesion polymerases, including the other member of the B-family, Pol. Recently, PrimPol, a primase-polymerase that is implicated to reinitiate replication after mitochondrial DNA damage, has been characterized to have low fidelity (54,55). It is interesting that we observe a prominent purine:purine mismatch in our experiments with Pol␣ (A:dA) as well as with PrimPol (G:dA).
High accuracy during the addition of the first dNMP to the RNA primer, which Pol␣ receives from primase during intramolecular transfer (6), might be important for a smooth switch from the RNA-to DNA-polymerizing mode of the primosome. The structure of the p180core/DNA:RNA/dCTP complex can provide a key to the effect of the 3Ј-ribose on Pol␣ fidelity. The 2Ј-OH modeled at the primer 3Ј-end has steric hindrance with the conserved Thr 1003 (Fig. 14C). The P 1 ribose shall move from Thr 1003 by ϳ0.4 Å to avoid steric clashing, which will affect the position of the T 1 -T Ϫ1 and can alter their ability to change the conformation upon mispair binding.
The structures of the catalytic domain of replicative DNA Pols provide a rationale for the mutagenic effect of substitutions of the conservative leucine (37,44,56,57), whose main chain interacts with the dNTP ␤-phosphate (Leu 864 in hPol␣; Fig. 4). This leucine is located in the hydrophobic pocket formed in hPol␣ by Phe 861 , Tyr 865 , Ile 868 , Ile 983 , and Leu 1036 . Bulky side chains of methionine and phenylalanine in place of Leu 864 cannot fit into this pocket, which would result in the distortion of the upper part of the ␣-helix (Fig. 4), affecting the position of Ser 863 , Leu 864 , and Tyr 865 . Among these residues, Tyr 865 is the most important for Pol selectivity. Such distortion of the dNTP-binding pocket can result in dNTP wobbling and reduced discrimination against mismatches.

The overall mechanism of RNA primer extension by Pol␣
Accumulated information from structural and functional studies allows us to understand the mechanism of chimeric RNA-DNA primer synthesis by the primosome, including the ability of Pol␣ to efficiently extend primers with varying DNA/ RNA ratio and bind double helices with different helical parameters. The following structural peculiarities of the Pol␣ catalytic domain are important to perform this task: small primer-binding interface, thumb flexibility, and the universal template:primer binding site, which is able to accommodate the substrates with different structures. Perhaps during evolution, Pol␣ developed the ability to fit the template:primer into a universal form of the double helix by deforming both DNA:DNA and DNA:RNA duplexes.
Structural studies of the human primosome point out that Pol␣ receives a 9-mer RNA primer from primase (6). The data presented here establish the four main steps of RNA primer extension by Pol␣: incorporation of the first dNMP (first step), second to fourth (second step), fifth to ninth (third step), and tenth and further (fourth step). The main features of the first step are the enhanced accuracy (Fig. 11) and the relatively low affinity for dNTP ( Table 3). The increased Pol␣ selectivity for a cognate dNTP is probably due to the presence of ribose at P 1 (Fig. 14C), whereas dNTP binding is also affected by the primer bending at P 4 (Fig. 7). The second step of RNA primer extension is characterized by reduced polymerization rate and increased processivity compared with Pol␣ activity on a DNA primer (Tables 5 and 6). For this step, primer bending at P 4 and the squeezing of a hybrid duplex between Gly 841 and Gly 1076 may affect template:primer translocation along the DNA binding cleft and dissociation from the polymerase. The third step of extension is similar to the fourth step, which corresponds to extension of a DNA primer ( Table 6). The hybrid part of the template:primer leaves the rigid and relatively narrow section of the DNA binding cleft and interacts only with the thumb, which is slightly more selective for a mixed AB-form of the double helix versus a B-form (Fig. 6, B and C). During the fourth step, the Pol␣ catalytic domain does not sense the hybrid part of a template:primer, and the characteristics of DNA synthesis should be identical to that on a DNA:DNA duplex.
The size of DNA tracks synthesized by Pol␣ on the leading and lagging strands has not been determined yet. It is possible that the length of chimeric RNA-DNA primers is precisely regulated, as was recently shown for the RNA primer (6). Only the structural properties of the primosome define the mechanism of RNA primer switching between primase and Pol␣, whereas additional factors should play a role in chimeric primer switching between Pol␣ and Pol⑀ or Pol␦ on the leading or lagging strand, respectively. In 2013, the mechanism of primer synthesis termination by Pol␣, based on its low activity in extension of homopolymeric DNA primers, had been proposed (16). As was shown later, the low activity of Pol␣ in primer synthesis on poly(dT) template is due to formation of unconventional DNA structures, which depends on the type of the divalent ion and its concentration (58). Moreover, comparable activity was observed during extension of DNA and RNA heteroprimers (59,60). The data presented here indicate that Pol␣ works on the DNA duplex with higher efficiency than on the hybrid one (Tables 4 and 5). On the other hand, Pol␣ has low processivity on the DNA duplex, which can favor its displacement from the template:primer by Pol⑀ on the leading strand and by the complex of the replication factor C and proliferating cell nuclear antigen on the lagging strand (replication factor C is then exchanged for Pol␦) (61).
It is possible that additional factors play a role in determination of the size of DNA track synthesized by Pol␣. For example, the primase accessory subunit (p58 in humans) has higher affinity for the template:primer (K d is 33 nM (62)) than Pol␣ (K d is 58 nM; Table 3) and can hold the 5Ј-end of RNA primer during its extension with dNTPs and promote Pol␣ reloading on the DNA primer (6). In this case, the linkers in primosome would allow Pol␣ to generate a ϳ20-mer DNA track. The primosome structure also predicts significant steric hindrance after synthesis of a ϳ10-mer DNA track due to ϳ360°rotation of the catalytic domain of Pol␣ relative to the C-terminal domain of p58; this would require a temporary dissociation of Pol␣ from the template:primer to untwist the linkers. In addition, it might be the signal from the replisome, leading to primosome dissociation from the template:primer. For example, as was shown in yeast, Pol␣ is tethered to the CMG helicase through the interaction between the N-terminal domain of the catalytic subunit and the replication factor Ctf4 (13). In humans, however, the Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes N-terminal domain of the accessory B-subunit connects Pol␣ to a replisome (63), similar to the mode of Pol⑀ recruitment to a replication fork (64). In both cases the CMG-interacting domain of Pol␣ is connected with the rest of the primosome by a long flexible linker (4). During Okazaki fragment synthesis, Pol␣ moves along the parental DNA strand in the direction opposite to the replication fork, and at some point, the tension in the linker, connecting the catalytic core with the replisome, will result in Pol␣ displacement from the primer 3Ј-end. Conducting structural and functional studies with the reconstituted eukaryotic replisome will shed more light on the mechanism of the template:primer switch from Pol␣ to Pol␦ and Pol⑀.

Oligonucleotides and reagents
Oligonucleotides were manufactured by IDT Inc. Deoxyribonucleotides used to extend the primers were obtained from Roche Life Sciences. Reagents for crystallization were obtained from Hampton Research. [␥-32 P]ATP was obtained from PerkinElmer Life Sciences.

Protein expression and purification
Cloning, expression, and purification to homogeneity of p180core (residues 335-1257), which contains the catalytic domain of hPol␣, have been described elsewhere (18). Peak fractions obtained from Heparin HP HiTrap column (GE Healthcare) were combined and dialyzed to the buffer specific for each application (see below).

Analysis of DNA-polymerizing activity in extension of fluorophore-labeled primers
Purified p180core was dialyzed to 10 mM Tris-HCl, pH 7.7, 100 mM KCl, 1% glycerol, and 1 mM DTT, and flash-frozen in aliquots. The primers had a Cy5 label at the 5Ј-end. Reactions were carried out at 35°C in buffer containing 30 mM Hepes-KOH, pH 7.8, 50 mM KCl, and 1 mM DTT. Reactions were started by mixing two solutions; one of them contained p180core and the template: primer, and the other one contained dNTP and the divalent metal (their concentrations are indicated in the figure legends; other components were the same in both stocks).

Crystallization
DNA:RNA and DNA:DNA duplexes were obtained at 0.2 mM concentration by annealing at 43°C for 30 min (after heating at 70°C for 1 min) in buffer containing 10 mM Tris-HCl, pH 7.9, and 70 mM KCl. In the case of DNA:RNA duplex, the sequences for DNA template and RNA primer were 5Ј-ATTACTATAG-GCGCTCCAGGC (the region complementary to a primer is underlined) and 5Ј-rGrCrCrUrGrGrArGrCrG/ddC/, respectively (/ddC/is a dideoxycytidine, which prevents polymerization during crystallization). In the case of DNA:DNA duplex, the template and primer were 5Ј-ATAGGCGCTCCAGGC and 5Ј-GCCTGGAGCG/ddC/, respectively. Dialyzed protein sample (15 M p180core) was diluted 1.5-fold with dialysis buffer (10 mM Tris-HCl, pH 7.7, 100 mM KCl, 1% glycerol, and 1 mM DTT) containing 36 M template:primer, 3.6 mM MgCl 2 , and 12 mM dCTP. The ternary complex was concentrated 10-fold and flash-frozen in liquid nitrogen. Before crystallization experiments, the aliquots of protein/template:primer/dCTP solutions were defrosted and centrifuged to remove the precipitate, and the sample monodispersity was verified with the dynamic light scattering. The screening of crystallization conditions was performed with the sitting-drop vapor diffusion method at 295 K by mixing 1 l of ternary complex solution with 1 l of reservoir solution. Initial screen solutions producing tiny crystals were optimized to produce well-shaped crystals at 295 K with reservoir solutions containing 0.8 mM zinc sulfate, 8.8% (v/v) PEG MME 550, and 50 mM MES, pH 6.5, for p180core/DNA:RNA/ dCTP and 0.8 mM cobalt chloride, 2 mM tris(2-carboxyethyl) phosphine (TCEP), pH 7.5, 250 mM 1,6-hexanediol, and 50 mM sodium acetate, pH 4.6, for p180core/DNA:DNA/dCTP.

Data collection
For data collection, each crystal was soaked in cryoprotectant solution (100 mM potassium chloride, 0.8 mM zinc sulfate, 1.1 mM magnesium chloride, 15% (v/v) PEG MME 550, 15% (v/v) ethylene glycol, and 50 mM MES, pH 6.5, for p180core/DNA: RNA/dCTP crystals, and 0.8 mM cobalt chloride, 2 mM TCEP, pH 7.5, 230 mM 1,6-hexanediol, 26% (v/v) ethylene glycol, and 50 mM sodium acetate, pH 4.6, for p180core-DNA:DNA-dCTP crystals) for a few seconds, scooped in a nylon-fiber loop, and flash-cooled in a dry nitrogen stream at 100 K. All initial diffraction data were obtained on a Rigaku R-AXIS IV imaging plate using Osmic VariMax TM HR mirror-focused CuK␣ radiation from a Rigaku FR-E rotating anode operated at 45 kV and 45 mA. Complete diffraction data sets were collected using synchrotron X-rays on the Argonne National Laboratory Advanced Photon source beamline 24-ID-E using ADSC Quantum 315 detector. All intensity data were indexed, integrated, and scaled with DENZO and SCALEPACK from the HKL-2000 program package (65). The crystals of p180core/DNA:RNA/ dCTP belong to trigonal space group P3 2 21 and diffract up to 2.2 Å resolution, and the crystals of p180core/DNA:DNA/ dCTP belong to tetragonal space group P4 2 2 1 1 and diffract up to 2.95 Å resolution. Both crystals contained only one copy of the ternary complex in the asymmetric unit. The crystal parameters and data-processing statistics are summarized in Table 1.

Crystal structure determination
Initial phases for p180core/DNA:RNA/dCTP structure were determined by the molecular replacement method using the coordinates of backbone atoms of the yeast Pol␦ catalytic core derived from its ternary complex with template:primer and dCTP (PDB code 3IAY) as a search model. The positions of magnesium and zinc ions were determined using an anomalous difference Fourier map. Molecular replacement and initial automated model rebuilding with Phenix (66) revealed over 70% of correctly built protein structure. The model building was continued and completed manually with Turbo-Frodo, and the structure was refined using standard protocols of CNS version 1.1 (67). After the addition of solvent molecules, the model was refined at 2.2 Å resolution to an R cryst of 21% and an R free of 23.9%. The structured region of protein starts with Glu 338 and ends with Thr 1244 . The electron density maps were poor or were missing also for the inter-Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes nal regions 674 -677, 809 -833, and 883-895; therefore, these regions were excluded from the structure.
The p180core/DNA:DNA/dCTP structure was determined by the molecular replacement method using the coordinates of p180core in the complex with DNA:RNA and aphidicolin (PDB code 5Q5V) as a search model. The model was adjusted manually with Turbo-Frodo. The positions of magnesium and cobalt ions were determined using an anomalous difference Fourier map. The structure was refined at 2.95 Å resolution using standard protocols of CNS version 1.1 (67) to an R cryst of 25.9% and an R free of 30.1%. The traceable electron density starts with Glu 338 and ends with Val 1248 . The regions 674 -677, 810 -833, and 883-895 with missing electron density were excluded from the model. The electron density was relatively weak for the portion of the thumb closer to the thumb tip, especially for residues 1137-1154, indicating their partial disorder and/or elevated mobility.
The refinement statistics for both structures are summarized in Table 1. The figures containing molecular structures were prepared with the PyMOL Molecular Graphics System (version 1.8, Schrödinger, LLC).

Analysis of DNA-binding activity
Purified p180core was dialyzed to 10 mM Tris-Hepes, pH 7.8, 200 mM NaCl, 2% glycerol, and 2 mM DTT, concentrated to 100 M, and flash-frozen in aliquots. The 15-mer primers (5Ј-GGACTCCGAGCTGCC; DNA and RNA) were labeled using [␥-32 P]ATP by T4 polynucleotide kinase from New England Biosciences. The 15-mer primers was were then annealed to the 25-mer template (5Ј-AATGTTTCATGGCAGCTCGGAG-TCC; the complementary region is underlined) at 95°C for 5 min, 55°C for 15 min, and 37°C for 10 min. The annealed substrates were prepared to contain 50,000 cpm/l and were used within approximately 1 week.
The p180core (4 M) was mixed with reaction buffer (10 mM Tris-HEPES (pH 7.8), 0.1 M NaCl, 5 mM MgCl 2 , 1 mM DTT) and 1 mg/ml BSA. Radiolabeled template:primer (6 nM, DNA or RNA primers) was mixed with the reaction buffer and 1 mg/ml BSA. Serial dilutions of the p180core solution were made so that solutions contained a range of 20 nM to 4 M enzyme. The diluted enzyme solutions were mixed with an equal volume of the template:primer solutions to obtain a final concentration of 10 nM to 2 M p180core and 3 nM template:primer. The solutions were allowed to incubate for 15 min at 25°C, and glycerol was added for a final percentage of 2%. The unbound and enzymebound radiolabeled template:primers were separated by 5% native PAGE in 0.5ϫ TBE for 33 min at 150 V at 4°C. The gels were visualized by the Molecular Imager FX phosphor imager (Bio-Rad) and quantified by Quantity One, version 4.6.9 (Bio-Rad).

Pre-steady-state kinetic assays
Pre-steady-state kinetic assays were performed using the RQF-3 rapid chemical quench apparatus (KinTek) at 37°C. All kinetics assays were carried out using reaction buffer. Pol␣ was incubated with template:primer and rapidly mixed with dNTPs before quenching with 0.5 M EDTA. Products were collected in a tube with formamide dye (0.1% bromphenol blue (w/v), 0.1% xylene cyanol (w/v)) and separated by denaturing urea PAGE. The radiolabeled products were visualized by the Molecular Imager FX phosphor imager (Bio-Rad) and quantified by Quantity One, version 4.6.9 (Bio-Rad).
For optimal separation of RNA primer-elongated products, a modified denaturing urea PAGE protocol was employed. A solution of acrylamide mix was made with the addition of 10% formamide (19% acrylamide, 1% bisacrylamide, 10% formamide, 7 M urea, 1ϫ TBE buffer (pH 8.3)). The acrylamide polymerization reaction was initiated with 0.05% ammonium persulfate and 0.1% TEMED. Before loading the gel with sample, the gel was run at 3000 V for at least 2 h to allow the gel to heat up. The elongated products containing equal parts of quenched product and formamide dye were then loaded onto the gel (Ͻ10 l) and allowed to run for ϳ6 h.

Burst incorporation kinetics
Burst incorporation kinetics were conducted with excess template:primer (9 M) to Pol␣ (3 M). To ensure that binding of the incoming nucleotide was not rate-limiting, a high concentration of dATP (200 M) was used in the polymerization reaction. The reaction products were quantified, and the concentration of product was plotted against time and fit to both a linear and a burst equation, where A is the burst phase amplitude corresponding to the active site concentration, k obs is the observed rate, k ss is the steady-state rate, and t is the time. In the event of misincorporation after a correct incorporation, 500 M dATP was used to ensure nucleotide saturation.

Single-and double-nucleotide incorporation assays
Single-and double-nucleotide incorporation assays were performed with excess Pol␣ (3 M) to template:primer (100 nM). In the case of single-nucleotide incorporation assays, the concentration of dATP was varied (1-300 M). For each experiment at a particular dATP concentration, the concentration of product was plotted against time and fit to a single-exponential equation, where A is the amplitude, k obs is the observed rate for the incorporation of the incoming dNTP, and t is the time. The k obs was then plotted against dATP concentration and then fit to a quadratic equation, to obtain the K d , the apparent binding constant for the incoming nucleotide for the Pol␣/template:primer binary complex, and k pol , the maximum rate of nucleotide incorporation.
In the case of double-nucleotide incorporation assays, Pol␣ first incorporates a dATP followed by a subsequent dTTP incorporation. The concentration of dATP was held constant at saturating amounts based on single-turnover experiments (100 M for DNA primer and 300 M for RNA primer), and the concentration of dTTP was varied (1-300 M). The data were fit using KinTek Global Explorer (35) to provide an estimate for the K d and k pol for the second incorporation event.

Interaction of human Pol␣ with DNA:DNA and DNA:RNA duplexes Processivity assays
Processive polymerization assays were performed under similar single-turnover conditions as the single-and doublenucleotide incorporation assays with excess Pol␣ (3 M) to template:primer (100 nM). The Pol␣/template:primer complex was mixed with saturating concentrations of all dNTPs (240 M each) to allow for complete extension of the primer substrate. The processivity data were modeled in KinTek Global Explorer to determine the k pol and k off , the rate of dissociation of the template: primer substrate from Pol␣, for each incorporation event.
In fitting the data to the models, the following assumptions were made: 1) all incorporation events were irreversible, 2) the K d and k pol values obtained from the single-incorporation experiments were manually fixed for the first incorporation in the double-incorporation modeling, 3) the k on values for template:primer binding to Pol␣ were assumed to be the same, disregarding primer length differences.

Data analysis and kinetic models
Analysis of the radiolabeled products through the phosphor imager provides raw data in count values. For each time point, the counts of each band present were summed to calculate the total amount of counts and subsequently the percentage of each product length present. These values were then normalized to add up to 100% and then converted into concentration amounts based on the total amount of nucleic acid substrate used in the assay.
The KinTek Global Explorer program requires user-inputted models to fit kinetic data for the estimation of rates (Fig. S2) (35,68). Briefly, the data from double-incorporation assays were fit to a two-step mechanism of dNTP binding followed by incorporation that was modified to account for two incorporations. The processivity data were fit using a model that allows for binding or dissociation of each DNA product and combines the dNTP binding and incorporation event into one rate.