DNA conformational changes at the primer-template junction regulate the fidelity of replication by DNA polymerase

Abstract

Local conformational changes in primer-template (P/T) DNA are involved in the selective incorporation of dNTP by DNA polymerases (DNAP). Here we use near UV CD and fluorescence spectra of pairs of base analogue probes, substituted either at the primer terminus or in the coding region of the template strand, to monitor and interpret conformational changes at and near the coding base of the template in P/T DNA complexes with Klenow fragment (KF) DNAP as the polymerase moves through the nucleotide addition cycle. Incoming dNTPs and rNTPs encounter binary complexes in which the 3′-end of the primer shuttles between the polymerization (pol) and exonuclease (exo) sites of DNAPs, even for perfectly complementary P/T DNA sequences. We have used spectral changes of probes inserted in both strands to monitor this two-state distribution and determine how it depends on the formation of ternary complexes with both complementary (“correct”) and noncomplementary (“incorrect”) NTPs and on the local sequence of the P/T DNA. The results show that the relative occupancy of the exo and pol sites is coupled to conformational changes in the P/T DNA of the complex that are partially regulated by the incoming NTP. We find that the coding base on the template strand is unperturbed by the binding of incorrect dNTPs, while binding of complementary rNTPs induces a novel template conformation. We conclude that, in addition to its editing function, primer strand occupancy of the 3′-exo site may also serve as a regulatory checkpoint for accurate dNTP selection in DNA synthesis.

Error rates during DNA synthesis for replicative and some repair DNA polymerases (DNAPs) are in the range of 1 in 10⁵ to 10⁸ (1, 2). While the inherent nucleotide selectivity that is thought to reflect Watson–Crick base pairing of the incoming dNTP with the templating base and the “steric fit” of the resulting base pair into the catalytic polymerase (pol) site of the DNAPs contribute significantly to maintain such extraordinary accuracy, fidelity is improved further by the 3′-exonuclease (exo) activity that removes mismatches from the growing primer terminus (2).Thus controlling the distribution of primer-template (P/T) DNA between the pol and exo sites of replicative DNAPs is essential for accurate DNA synthesis. Here we report evidence for a role for this distribution in regulating replication fidelity that transcends the editing function in KF of Escherichia coli DNAP I.

Structures of binary DNAP-P/T and ternary DNAP-P/T-dNTP complexes have been obtained for a variety of A-family DNA polymerases (3–6). All share a common architecture resembling a “right hand”: including a “thumb” subdomain that binds P/T DNA, a “fingers” domain that interacts with the incoming dNTP, and a “palm” domain containing the highly conserved residues of the polymerase active site (5). Until recently it was thought that DNAPs cycle between two distinct conformational states during DNA synthesis: a binary “open” conformation in which the fingers are farther away from the active site and the P/T DNA (shown schematically* in Fig. 5C) and a ternary “closed” conformation in which the fingers close around the P/T DNA and the incoming dNTP to form the catalytically active complex (Fig. 5E) (5, 6). Recent studies, including this one, suggest that structurally distinct and mechanistically important intermediate states exist as well.

Fig. 5.

A schematic overview of the disposition and occupancy of the active sites of DNA polymerase in binary (A, B, and C) and ternary (D and E) complexes. The figures for the open binary pol (5C) and closed ternary complexes (5E) are adapted from ref. 5, while that of the open binary exo complex (5A) is based on the structure in ref. 3. Cartoons representing the conformations of the various KF-DNA complexes in the reaction are shown at the top of each panel, and blow-ups of the conformations of the DNA bases are shown just underneath. KF is shown in green, while the polymerase subsites are color coded as follows: yellow (insertion site); gray (preinsertion site); pink (intermediate site between pol and exo active sites); and cyan (exo site). T, F, and P represent the thumb, fingers, and palm subdomains, respectively. The primer and the template DNA strands are shown in red and blue, respectively, and the coding base (n) is shown in black. The incoming dNTP/rNTP is shown in purple. In the open binary complex (5C), n is bound at the preinsertion site (between the O and O1 helices) and a conserved tyrosine (Tyr, orange) blocks access to the insertion site (shown in yellow). Formation of the closed conformation (5E) in the presence of the complementary dNTP (purple) involves rearrangement of the O and O1 helices, which simultaneously blocks the template preinsertion site and unblocks the insertion site. These rearrangements move the coding base (n) of the template to the insertion site, where it pairs with an incoming dNTP. Incoming dNTP occupies the intermediate preinsertion site (5D) in a conformation previously proposed for T7 RNAP (12, 15).

Crystallographic studies (3, 5) have shown that the 3′-end of the primer strand of P/T DNA can bind to and occupy either of the catalytic sites (which are separated by about 30 Å), forming pol or exo complexes. In addition three distinct conformations have been observed for the coding base, n (5) in DNAP. Thus n may occupy the “preinsertion” site in the open binary complex (Fig. 5C); this is a pocket between the conserved O and O1 helices of the fingers domain. When bound in this pocket the n base is “flipped out” of the DNA helical axis and is not accessible to the incoming dNTP. n can also occupy the “insertion” site in the closed ternary complex prior to catalysis (Fig. 5E); here the n base is stacked next to the 3′ primer terminus and forms a Watson–Crick base pair with the incoming dNTP. Significant global motions of the O-helices are required to transfer the n base from the preinsertion site to the insertion site (5). Finally, following phosphodiester bond formation and DNA translocation, n occupies the “postinsertion” site, while the newly extended 3′ primer terminus occupies the pol active site in preparation for the next round of synthesis.

Recent kinetic studies have suggested that a conformational change prior to “finger-closing” may be involved in an early checkpoint for correct dNTP incorporation (7–10). The conformational changes in the DNAP or the P/T DNA that might be involved in this checkpoint have not been defined. In the homologous T7 RNA polymerase (RNAP) the coding base in a preinsertion site appears to hydrogen bond with the incoming complementary rNTP, thereby forming an “open ternary” complex (11, 12). A similar conformation has been proposed, though not observed, to explain the kinetic selection of dNTP by DNAP (13–15). Following Waksman (15), we refer to such a positioning of the coding base as an “intermediate preinsertion” site conformation (Fig. 5D), to distinguish it from the preinsertion site conformation in the binary DNAP complex (Fig. 5C) in which the coding base is not accessible to incoming dNTPs (5).

To investigate the structural origins of this kinetic checkpoint we have monitored local conformational changes within the P/T DNA during dNTP selection by means of fluorescence and near UV circular dichroism (CD) spectra changes for adjacent pairs of 2-aminopurine (2-AP, an adenine analogue) (11) or 6-methyl isoxanthopterin (6-MI, an analogue for guanine) (16) bases at specific positions within the P/T DNA. Previously we used this technique to map local conformations at specific positions within the primer DNA bound at either the pol or the exo active site of KF, and to measure the distribution of the primer terminus between these two enzymatic sites (17). Here we probe the local conformations of both the primer terminus and the template n base bound to KF in the presence of various potential dNTP and rNTP substrates to further define structural aspects of possible checkpoint mechanisms in DNAP fidelity control.

Results

Using the Near UV CD Spectra of 6-MI Dimers to Probe DNA Conformation Changes at the Coding Base Within Polymerase-P/T DNA Complexes.

We introduce a novel base analogue probe, the 6-MI dimer (two adjacent 6-MI bases on the same DNA strand), to monitor the local conformations of the coding base of P/T DNA. The fluorescence properties of single 6-MI (Fig. 1) bases have been used previously to study the behavior of guanine bases within DNA (16). The structure of the bp formed by 6-MI with cytosine is shown in Fig. 1. We note that the 6-MI dimer has a significantly larger low-energy CD signal per mole residue (Fig. 1B) than do the 2-AP and pyrrolo C (PC, a cytosine analogue) probes that we used in our earlier studies (11, 18).

Fig. 1.

CD spectra of 6-MI modified single-stranded DNA and duplex DNA–DNA and RNA–DNA constructs. The molecular structures for the canonical G∶C bp and the 6-MI∶C bp are shown at the top. The P/T DNA construct used is shown above the graphs and the positions of the 6-MI dimer probes are indicated as “gg”. n denotes the position of the template coding base, while the positive and negative numbers following n represent bases positioned either downstream or upstream from the n base, respectively. The 16-mer primer (top) strand was either ssDNA, forming a DNA–DNA duplex (red) with the 6-MI modified template (bottom) strand (dark green), or RNA, resulting in an RNA–DNA heteroduplex (dark blue) of the same sequence. Panel A shows the high energy CD spectra of the constructs that are dominated by the canonical bases. Panel B shows the low-energy CD spectra of the 6-MI analogue dimers within the constructs. The solid lines represent 6-MI modified constructs while the dashed lines denote the corresponding unmodified oligonucleotides with identical sequence.

Fig. 1 A and B show characteristic CD spectra, at wavelengths below or above 300 nm, of the 6-MI dimer (denoted as gg) in which the probes are either unpaired and partially stacked in single-stranded (ss) DNA, or fully base-paired and stacked in either double-stranded (ds) DNA or an RNA-DNA hybrid within a P/T construct. A bimodal signal centered at the absorption maximum is characteristic of reporter bases stacked in the B-form dsDNA conformation, while partially unstacked probe conformations display a positive or negative CD band, depending on their chiral environment (11). The CD signal of the 6-MI dimer is centered in the 320–390 nm wavelength region of the near UV spectrum and is therefore well removed from the larger global signals below 300 nm that arise from the canonical bases and bps. We note that the 230–300 nm CD spectra are essentially unchanged by the substitution of 6-MI bases for guanine, showing that the 6-MI dimer probe fits well into the cooperative duplex conformations of both dsDNA and RNA-DNA hybrids (compare dotted and solid spectra in Fig. 1A). Somewhat larger differences were observed in the ssDNA spectra, suggesting that the noncooperative stacking of adjacent 6-MI probes in ssDNA may differ somewhat from the stacking of pairs of G bases when the conformation is not “cooperatively locked in” by a fully duplex structure. Additional characterization of 6-MI dimer probes as GG replacements (Fig. S1) is summarized in SI Text and will also be published elsewhere in more extended form (Datta et al., ms in preparation).

Distribution of the P/T DNA Primer Strand Between the pol and exo Sites of KF in the Binary Complex.

Using 2-AP dimer probes in the primer strand as a monitor, we have previously established experimental conditions that form “end-state” binary KF-P/T DNA complexes in which the 3′-end of the primer strand is bound entirely in either the pol or the exo site (17). Here 6-MI dimer probes located at positions n,n + 1 are used to characterize template strand conformations at the n base in these end-state complexes. Fig. 2C shows CD spectra for the construct and the complex in Ca²⁺ buffer, in which the 3′-primer terminus of a P/T DNA construct with three terminal mismatches is bound exclusively at the exo site. Fig. 2B shows spectra in EDTA buffer (containing 2 mM EDTA and no added divalent cations), in which the primer of a fully base-paired P/T DNA construct is totally bound in the pol site (see ref. 17 and SI Text). In contrast, and as shown previously with 2-AP dimer probes positioned at the end of the primer strand, the 3′-ends of fully complementary P/T DNA constructs in Ca²⁺ buffer are distributed between the two active sites (Fig. 2A).

Fig. 2.

CD spectra of 6-MI labeled P/T DNA constructs in binary complexes with KF DNAP. Low-energy CD spectra of the KF-DNA binary complex with (A) the upper construct in Ca²⁺ buffer; (B) the upper construct in EDTA buffer; and (C) the lower (three terminal mismatches) construct in Ca²⁺ buffer.

With the primer terminus fully bound in the exo site, these results show that the CD signal for this template-substituted complex resembles closely that obtained with the free DNA construct (Fig. 2C), suggesting that in the binary exo complex the local conformation of the n,n + 1 template bases are unperturbed relative to the free construct. In contrast, with the primer strand bound in the pol site, the CD spectrum of the template-substituted dimer probe looks very different, exhibiting a peak near 360 nm (Fig. 2B, pink trace) that suggests that the 6-MI probes at the coding base are significantly unstacked under these conditions (11). This result doubtless reflects the local distortion of the template strand that is required to place the coding (n) base in the preinsertion site of the binary DNAP pol complex (5) and confirms that this distorted conformation also forms in solution at the template base of a fully complementary binary DNAP-P/T DNA complex.

These results demonstrate unequivocally that the template bases at and near coding base n assume very different local conformations when the primer terminus is bound in the pol or the exo site of KF, providing important insight into how both strands of the P/T DNA are handled by KF during both synthesis and editing. Furthermore, Fig. 2A shows that the CD signal of 6-MI residues in a matched P/T DNA construct in the presence of divalent Ca²⁺ appears to be a linear combination of these spectra (pink trace), showing that the two-state character of the pol–exo distribution can be established and monitored by using probes in either strand. We quantified this equilibrium distribution by deconvoluting the CD spectrum of the 6-MI dimer in the template strand as previously described for 2-AP dimer probes placed at the 3′-end of the primer strand (17).

Experiments using P/T DNA constructs containing either GG (or gg) at the n,n + 1 template positions or AA (or 2-AP dimer, denoted as aa) at the 3′-end of the primer gave identical 70% exo to 30% pol distributions (solid black fits; Fig. 3 A and B). Hence, as long as the overall sequence of the P/T DNA was not changed, these different analogue probes at these positions yielded the same distribution of primer ends between the exo and pol sites, again confirming that these base analogues can effectively replace their canonical counterparts in a variety of contexts.

Fig. 3.

Measuring the equilibrium distribution of the primer terminus between the pol and exo sites in binary complexes using dimer probes placed either at the primer terminus or at the coding base. Low-energy CD spectra for KF-P/T DNA complexes containing either (A) 2-AP or (B) 6-MI probes at the indicated positions. The constructs used are above each panel, and the positions of the 2-AP and 6-MI dimer probes are indicated as “aa” and “gg”, respectively. n_(A), n_(T), n_(G) and n_(g) identify the n base (A,T,G and 6-MI, respectively) in the various constructs. The black traces represent the best fits of the corresponding experimental curve to a linear combination of the exo-mode and pol-mode spectra. Inset in A shows the CD spectra of the free primer-labeled constructs and demonstrates that these spectra are not altered by changing the n base.

In contrast, changing the sequence of the template DNA immediately downstream of the P/T junction does significantly influence this distribution, with AA or TA bases in the n,n + 1 positions decreasing the fraction of primer ends in the exo site to 40%, compared with 70% exo for GG bases in the same positions (Fig. 3A). This result may argue that “stiffer” sequences containing GG (or gg) at the n,n + 1 positions in the template strand are more difficult to distort in placing the n base into the preinsertion site, thus favoring a higher population of primer ends in the exo site for GG sequences relative to sequences containing AA or TA in these positions. We note that all the P/T DNA constructs containing the 2-AP dimer probe with varied bases in the n and n + 1 template positions displayed the same low-energy CD spectra in the absence of KF (Fig. 3A, Inset), demonstrating that the changed spectral signals observed in the binary complexes result from different conformations of the P/T DNA construct that depend on the binding of the primer terminus to the two different catalytic sites of KF DNAP. We expect (see also ref. 17) that sequence alterations elsewhere in the P/T DNA construct will also perturb the two-state pol to exo primer binding distribution.

Local DNA Conformations Within Ternary Complexes.

We have also monitored conformational changes in the P/T DNA framework induced by adding NTP substrates to binary complexes to form ternary (KF-DNA-NTP) complexes. The addition of the correct dNTP (dCTP) to the binary complex provoked a large increase in the depth of the CD trough near 340 nm, with similar spectra being obtained for DNA constructs with either a 3′-deoxy (3′-H) or a 3′-OH nucleotide residue positioned at the 3′-primer end (see Fig. 4 A and B). The use of a nonextendable primer allowed us to capture the CD signal of bases n,n + 1 within a closed and catalytically active DNAP complex, where coding base n presumably occupies the insertion site (Fig. 5E) (5, 6). The addition of rCTP (the correct ribo-NTP) resulted in a much smaller change in the CD signal (Fig. 4A), although the direction of change was the same as that observed with dCTP. The addition of dCTP to a binary complex formed with the 3′-OH P/T construct resulted in the gradual appearance of a bimodal signal with a peak at 380 nm (Fig. 4B, Inset, and Fig. S2), while the spectrum of the construct containing the 3′-H primer (Fig. 4A) remained unchanged under identical conditions. Furthermore, no time-dependent changes in the CD spectrum of the n,n + 1 bases relative to the binary complex were observed in the presence of incorrect dNTPs or rNTPs (Fig. S2). Hence, the change in CD signal with dCTP as a function of time must reflect the slow incorporation of cytosine. The use of Ca²⁺ rather than Mg²⁺ in these experiments slows the reaction rate (7) sufficiently to permit time-dependent monitoring of the chain extension process.

Fig. 4.

Near UV CD spectra of the n,n + 1 and n - 1,n template base positions during dNTP selection. Low-energy CD spectra of KF-P/T binary and KF-P/T-NTP ternary complexes with P/T constructs containing: (A) a nonextendable 3′-deoxy primer terminus with 6-MI at the n,n + 1 positions; (B) an extendable primer terminus with 6-MI at n,n + 1 positions; and (C) an extendable primer terminus with 6-MI at n - 1,n positions. Insets show time-dependent changes in the low-energy CD in the presence of complementary dCTP, due to its incorporation at the primer terminus. Gels show the products of primer-extension reactions with either (D) the nonextendable or (E) the extendable constructs with 6-MI at the n,n + 1 positions under the same reaction conditions used for the CD experiments. Quantitative estimates of the products formed with the extendable primer after 60 min of reaction are also shown.

Unlike complementary dCTP or rCTP, the incorrect dNTPs and rNTPs did not induce any significant rearrangement of the n,n + 1 bases in the template (Fig. 4A and B). Furthermore, primer extension experiments using radioactively labeled primer DNA with a 3′-OH terminus, performed under the same conditions used for the spectroscopic measurements, showed negligible misincorporation in the presence of noncomplementary dNTPs and rNTPs (Fig. 4E). Hence our spectroscopic measurements reveal P/T conformational changes that appear to correlate with discrimination between correct and incorrect dNTPs and rNTPs by the KF-P/T DNA complex.

Additional experiments were performed to probe the n - 1,n and n + 1,n + 2 template positions in binary and ternary KF complexes (Fig. 4C and Fig. S3). The addition of dCTP (the correct dNTP) resulted in the characteristic trough at 340 nm in the CD spectrum of probe residues at the n - 1,n template positions. This signal is similar to that observed for the ternary complex with the n,n + 1 construct in the presence of the correct dNTP (Fig. 4A and B). Due to the extendable nature of the primer strand used for probing the n - 1,n positions, we also observed the time-dependent evolution of the CD spectrum, reflecting the slow incorporation of cytosine at the primer terminus (Fig. 4C, Inset). The characteristic trough was not observed with the n + 1,n + 2 construct when the correct dNTP was added to this KF-P/T DNA complex (Fig. S3). Therefore, the CD trough at 340 nm observed for constructs with probes in the n - 1,n and n,n + 1 template positions must reflect a local conformation in which the n base occupies the active site in a Watson–Crick base-paired state with the correct dNTP within the KF closed ternary complex. The ternary complex involving the n - 1,n construct with added incorrect dNTPs showed significant unstacking of the probe bases relative to the binary complex (Fig. 4C), while no difference was observed with the n,n + 1 construct (Fig. 4A and B).

Complementary fluorescence and CD experiments were also performed with constructs containing 2-AP dimer probes at the primer terminus to monitor the effects of NTPs on base pairing at the duplex terminus of P/T DNA. The results were fully supportive of the data presented above and are presented in SI Text, including Fig. S4 and Fig. S5.

Discussion

Accurate replication is accompanied by local conformational changes of the DNA framework of DNA-DNAP complexes. We monitor these changes by taking advantage of relationships between local nucleic acid conformations and the low-energy CD spectra of base analogue probes placed at specific positions within the P/T DNA. Using the 2-AP dimer probe we have recently (17) shown in solution that the terminal bases of the primer strand bind in an extended conformation in the exo site, while remaining stacked when bound at the pol site of KF, as expected from crystal structures of homologous A-family DNA polymerases (3, 6). The low-energy CD spectra of the 6-MI dimer also provide characteristic signals for stacked and unstacked bases (Fig. 1) and extends the repertoire of dimer probes that can be used to monitor local DNA conformations (11, 18). Here we report DNA conformational changes associated with dNTP selection, using P/T constructs carrying 2-AP dimer probes at the 3′-end of the primer strand or 6-MI dimer probes located at or near the coding base position (n) of the template strand.

In addition to the conformational changes induced in the primer strand, our results show that the conformation of the template strand in the binary KF-P/T DNA complex is also significantly changed when the primer strand terminus moves from the pol to the exo site (Figs. 2 and 3). The template strand bases at positions n and n + 1 are largely unstacked (CD peak at 360 nm) when the primer is bound at the pol site, consistent with assumed positioning of template coding base n in the preinsertion site in this complex (Fig. 5C). In contrast, when the primer terminus is bound in the exo site (Fig. 5A) the bases at positions n,n + 1 of the template strand are stacked, showing a bimodal (exciton-coupled) CD spectrum. In this configuration the template strand may interact with the RRRY motif (19), a recently postulated template binding sequence in KF that is physically separate from the pol site (5).

DNA Conformations in Ternary Complexes Respond to NTP Binding.

We [and others (17, 20, 21)] have observed that under most conditions the primer strand at P/T DNA junctions populates both the pol and exo sites of binary DNAP complexes, even in the absence of a terminal mismatch. DNA binding in the exo site is usually thought to be involved in the replication pathway only after covalent misincorporation; that is, in the context of primer editing to remove an incorrect 3′ terminal base. However, because incoming NTPs encounter binary complexes in which P/T DNA is in a dynamic equilibrium between the pol and exo binding sites, it is possible that this equilibrium could participate in an early structural checkpoint for NTP selection.

The intense CD trough near 340 nm was observed only in the presence of the correct dNTP (Fig. 4), which stabilizes the closed conformation of the ternary DNAP complex with the template n base in the insertion site and base-paired with the incoming dCTP (Fig. 5E). In this conformation the primer strand forms a duplex with the template DNA (Fig. S5), thereby positioning the α-phosphate of the incoming dNTP for an in-line nucleophilic attack on the 3′-OH of the primer terminus (2). We find no indication of the formation of a closed ternary complex in the presence of incorrect dNTPs (Fig. 4), in agreement with published kinetic studies (7, 14, 22).

The CD spectrum of the template n,n + 1 bases in the presence of noncomplementary dNTPs is different from that observed in the presence of either dCTP or rCTP (Fig. 4A). In ternary complexes with noncomplementary dNTPs, the probe residues at positions n - 1,n become more unstacked (Fig. 4C), which could reflect sequestration of n base into the preinsertion site. However, probes positioned at n,n + 1 did not exhibit any significant change relative to the respective binary complexes (Fig. 4 A and B), suggesting that bases at these positions display the same conformational equilibrium with the pol site as in the binary complex. This apparent discrepancy may reflect the extent of primer shuttling into the pol site for the two constructs. The terminal bp at the P/T junction (position n - 1) is G•C in the construct in which the n - 1,n positions were probed and A•T for the construct used for probing the template n,n + 1 positions, suggesting the G•C bp might stabilize duplex DNA bound in the pol mode more than A•T. Finally, although the template n,n + 1 bases do not show any change in the equilibrium distribution in the presence of incorrect dNTP, 2-AP fluorescence quenching indicates that the primer terminus of the same construct may partially shift toward the pol site (Fig. S5 A and B). In this hypothesis, binding of incorrect dNTP could cause less stable P/T junctions (terminal A•T bp) to assume a conformation where the P/T DNA has moved partway toward the pol site, perhaps without complete sequestration of the n base into the preinsertion site (Fig. 5B). The presence of such an intermediate transition state at the “crossroads” between replication and editing has been suggested earlier for other proof-reading-proficient DNAPs (23, 24). Here we report the possible utilization of such a transient state in the nucleotide selection pathway of DNA replication.

The presence of saturating concentrations of the correct ribo-NTP (rCTP) also significantly quenches the 2-AP dimer probe, consistent with the transfer of the primer terminus from the exo site, where it is unwound, to a duplex with template DNA. However, this transfer occurs only at much higher rCTP concentrations than required with dCTP (Fig. S5). The template n,n + 1 bases also stack more on rCTP binding, although the CD signal is not characteristic of the closed complex formed with correct dCTP (trough at 340 nm). The CD spectra of the ternary complex with rCTP could have a small component of signal from the closed complex. Alternatively, the binding of rCTP could drive the n,n + 1 bases into an arrangement such as the intermediate preinsertion configuration (Fig. 5D), which could contribute such a signal. In either case complementary rNTP binding moves the coding base from the preinsertion site, most likely to a position where it can make H-bond contact with the n base but it cannot move the n base into the insertion site (compare brown and blue traces Fig. 4A). It has been previously reported that complementary rNTP inhibits finger-closing within the complex (7), perhaps due to a clash of the conserved “steric gate” residues with the ribose sugar of the rNTP (25). Here we show directly that the coding base does not transfer to the insertion site in the presence of complementary rNTP.

Possible Structural Role of the exo–pol Site Distribution of the 3′ Primer Terminus as a Fidelity Checkpoint in DNA Replication.

Taken together these results suggest that initial discrimination against incorrect nucleotides by KF is achieved in two steps, both of which precede the final finger-closing process. NTPs first encounter the binary DNAP complex and modulate the partitioning of the P/T DNA to the pol site. This partitioning is also dependent on the stability of the bp at the P/T junction, leading to selective stabilization of the coding base at the preinsertion site. The dynamics of the partitioning of the primer strand between the pol and exo active sites has been studied extensively in the context of proofreading. However, the regulatory role of NTPs in these events remains to be examined (see SI Text). Here we show that the binding of various NTPs promotes the depopulation of the exo site in the following order of increasing effectiveness: incorrect dNTPs or rNTPs < correct rNTP < correct dNTP (Fig. S5), as would be expected if partitioning of the primer terminus between the exo and pol sites were coupled to NTP selection. Binding of complementary rNTP also modifies the coding base conformation (Fig. 4A), while incorrect dNTPs do not have this effect, indicating that they may be discriminated against in the preinsertion site prior to this conformational change. As argued above, the CD spectral changes that accompany rCTP binding could be explained by formation of a complex with the template coding base in an intermediate preinsertion site (Fig. 5D), where it could pair with the complementary rNTP as proposed for RNAP (12). These changes in P/T DNA conformation are doubtless accompanied by conformational changes in the polymerase, which may also be involved in an early kinetic checkpoint for dNTP selection (7, 15, 22).

These results suggest that the role of the exo site goes considerably beyond its enzymatic proofreading function. Binding of the primer in the exo site partially unwinds the DNA duplex and could thus block the translocation of the replication complex, regulating the advance of the polymerase. For example, interactions between PCNA and DNAP from Pyrococcus furiousos appear to decrease the binding of primer strands at the exo site and stabilize the P/T DNA duplex within the polymerase, thereby favoring processive elongation (26). In addition, partitioning of the primer terminus between the polymerase and exonuclease active sites modulates translesion synthesis by E. coli DNA Pol II (27). Polymerase features that are involved in control of the occupancy of the exo site are beginning to be investigated by protein engineering (27). Not all DNA polymerases have an exo binding mode; repair polymerases in the X and Y families and certain A-family polymerases that perform translesion and mutagenic DNA synthesis lack proofreading exonuclease activity (28). It is possible that primitive replicative DNAPs may have stabilized unwound P/T DNA as a means of placing replication under the control of transacting protein or NTP components as shown here, and that such regulatory mechanisms subsequently evolved to produce 3^′ → 5^′ exonuclease activity and replicative editing.

Materials and Methods

Materials.

Unlabeled DNA oligonucleotides were purchased from Integrated DNA Technologies (Coralville, IA). 6-MI-labeled DNA oligonucleotides were from Fidelity Systems (Gaithersburg, MD), while 2-AP modified oligonucleotides were purchased from Operon (Huntsville, AL). P/T DNA constructs were prepared as described (17). The clone (plasmid pXS106) for the exo^- D424A derivative of KF, and the host (CJ 376) cells were gifts from Catherine Joyce (Yale University). The protein was expressed and purified as described previously (17). Unless otherwise stated, all experiments were performed at 25 °C in Ca²⁺ buffer (20 mM Hepes (pH 7.9), 50 mM sodium acetate, 5 mM Ca(OAc)₂, and 1 mM DTT) at equimolar (3 μM) concentration of DNA and KF and 0.5–2 mM concentrations of dNTP and rNTP substrates.

Spectroscopic Procedures.

The fluorescence and CD spectra were measured as described in SI Text and (17). The CD spectra shown are reported as ε_L-ε_R in units of M^-1 cm^-1 per mole of 6-MI (for 6-MI dimer probes between 300 and 450 nm) or per mole of nucleotide residues (for canonical DNA bases between 230 and 300 nm).

Polymerase activity assays.

Polymerase activities were measured with P/T DNA constructs labeled at the 5′-end with γ-P³²-ATP and the DNA products were analyzed as described in SI Text and (17).