Crystal structures of a ddATP-, ddTTP-, ddCTP, and ddGTP- trapped ternary complex of Klentaq1: Insights into nucleotide incorporation and selectivity

Abstract

The mechanism by which DNA polymerase I enzymes function has been the subject of extensive biochemical and structural studies. We previously determined the structure of a ternary complex of the large fragment of DNA polymerase I from Thermus aquaticus (Klentaq1) bound to a primer/template DNA and a dideoxycytidine 5′-triphosphate (ddCTP). In this report, we present the details of the 2.3-Å resolution crystal structures of three additional ternary complexes of Klentaq1 bound to a primer/template DNA and a dideoxyguanosine 5′-triphosphate (ddGTP), a dideoxythymidine 5′-triphosphate (ddTTP), or a dideoxyadenosine 5′-triphosphate (ddATP). Comparison of the active site of the four ternary complexes reveals that the protein residues around the nascent base pair (that formed between the incoming dideoxynucleoside triphosphate [ddNTP] and the template base) form a snug binding pocket into which only a correct Watson-Crick base pair can fit. Except in the ternary complex bound to dideoxyguanosine 5′-triphosphate, there are no sequence specific contacts between the protein side chains and the nascent base pair, suggesting that steric constraints imposed by the protein onto the nascent base pair is the major contributor to nucleotide selectivity at the polymerase active site. The protein around the polymerase active site also shows plasticity, which may be responsible for the substrate diversity of the enzyme. Two conserved side chains, Q754 and R573, form hydrogen bonds with the N3 atom in the purine base and O2 atom in the pyrimidine base at the minor groove side of the base pair formed by the incorporated ddNMP and the corresponding template base in all the four ternary complexes. These hydrogen-bonding interactions may provide a means of detecting misincorporation at this position.

Faithful replication of DNA molecules by DNA polymerases is essential for genome integrity and stable transmission of genetic information in all living organisms. DNA replication in vivo is accomplished with remarkable accuracy (error frequency of ∼10⁻⁸ to 10⁻¹⁰ [Echols and Goodman 1991]). At least three distinct processes contribute to the high fidelity of DNA replication: (1) the DNA polymerization reaction itself has a very low error frequency (∼10⁻³ to 10⁻⁵); (2) the proofreading reaction reduces the error frequency by one or two orders of magnitude; (3) the postreplication repair system further increases DNA replication fidelity by approximately three orders of magnitude. Understanding the mechanisms by which DNA polymerases function in these processes to achieve high-fidelity DNA synthesis is one of the major goals of current biological studies on DNA polymerases.

Bacterial DNA polymerase I (Pol I) enzymes have served as model systems for studying the mechanisms of polymerase function, especially that of the DNA polymerization reaction. These enzymes are characterized by multidomain structures, which contain a polymerase domain, a 3′ to 5′ exonuclease domain, and a 5′ to 3′ exonuclease domain. Limited proteolysis of the Pol I enzyme from Escherichia coli results in two distinct fragments: the large C-terminal fragment (Klenow fragment) contains the polymerase activity and the 3′ to 5′ exonuclease activity; the small N-terminal fragment contains the 5′ to 3′ exonuclease activity (Klenow et al. 1971). The crystal structure of the Klenow fragment from E. coli Pol I shows that the polymerase domain has a shape reminiscent of a right hand, with a large cleft formed by the fingers, thumb and palm domains (Ollis et al. 1985). Comparison of the structures of DNA polymerases from different organisms shows that the hand-shape architecture is a common feature of the polymerase domain among all known polymerases, although the sequences and structural details of these enzymes are widely diverse (Steitz 1999).

A minimal kinetic scheme (Johnson 1993) for the mechanism of nucleotide incorporation by Pol I enzymes has been established, based on studies using E. coli Pol I or its Klenow fragment (Kuchta et al. 1987, 1988) and phage T7 DNA polymerase (Patel et al. 1991; Wong et al. 1991). In this kinetic scheme, the enzyme–DNA complex exists in two states: the open state and the closed state. The nucleoside 5′-triphosphate (dNTP) substrate binds to the enzyme–DNA complex in the open state. A subsequent rate-limiting conformational change in the enzyme turns the enzyme from the open state to the closed state. In the closed state, the chemical reaction occurs and the nucleotide is incorporated. A second rate-limiting conformational change takes place after the chemical reaction, and the enzyme returns to its open state. Pyrophosphate release and DNA translocation may occur at this step, and the enzyme is ready for the next round of nucleotide incorporation. The structural basis for this open to closed transition of the enzyme has been revealed recently based on the crystal structure of the quaternary complex of T7 polymerase bound to a primer/template DNA, a ddGTP, and thioredoxin (Doublié et al. 1998) and that of the ternary complex of the Klentaq1 polymerase (i.e., the large fragment of DNA polymerase I from Thermus aquaticus) bound to a primer/template DNA and a ddCTP (Li et al. 1998b). These structures describe the closed state of the enzyme–DNA complex after ddNTP binding. In these structures, the fingers domain of the polymerase has a different conformation from that seen in other known polymerase structures (including the apo and nucleotide- or primer/template DNA-bound Klentaq1; Korolev et al. 1995; Li et al. 1998a, 1998b). The tip of the fingers domain undergoes an inward rotation by ∼40°–46°, toward the polymerase active site at the base of the palm domain. This conformational change results in a closing up of the cleft formed by the fingers, thumb, and palm domains, and consequently a tight binding complex with substrates properly positioned for catalysis is formed.

In both the closed Klentaq1 and T7 complexes, the side chains of the protein form a narrow pocket around the base pair between the incoming ddNTP and the template. Because there are no sequence specific contacts between the protein and the incoming ddNTP base, it was hypothesized that the snug fit of the protein around this nascent base pair is responsible for the selection of the correct nucleotide by the polymerase. This steric selection hypothesis remains to be tested further.

In this study, we describe the details of the 2.3-Å resolution crystal structures of three additional closed ternary complexes of Klentaq1 bound to a primer/template DNA and a ddATP, a ddTTP, or a ddGTP. An initial report has been published in Li et al. (1999); this report was restricted to the discovery of a conformational change in the side chain of one residue, R660, which, on incorporation of ddGTP, is directed toward the incorporated nucleotide and makes tight hydrogen-bonding interactions with the O6 and N7 atoms of the G base. This localized conformational change was shown to be responsible for higher rates of ddGTP incorporation by Klentaq1. However, by providing a view of the active site conformation during incorporation of each individual nucleotide, this unique set of ternary complex structures provides insights into the mechanism of nucleotide selectivity during the DNA polymerization reaction. These were not described in the report by Li et al. (1999) and are described below.

Results and Discussion

Description of overall structures

The closed ternary complexes of Klentaq1 bound to a primer/template DNA and a ddGTP, a ddATP, or a ddTTP (defined as ddGTP-, ddATP-, or ddTTP-trapped complex, respectively) crystallized in conditions similar to those used previously to crystallize the Klentaq1–DNA–ddCTP complex (defined as ddCTP-trapped complex [Li et al. 1998b, 1999]). The crystals diffracted to the same resolution (2.3 Å) with same space group and similar cell dimensions as those of the ddCTP-trapped complex (Li et al. 1999). The protein in each of the complex structures contains residues 293 to 831. The first 19 residues (residues 274–292) and the C-terminal residue (residue 832) in the construct were disordered and not included in the model. The first two single stranded nucleotides at the 5′ end of the DNA template (A_T1 and A_T2 in nomenclature adopted in Li et al. [1998b]) were not included in the model because of lack of interpretable electron density. An incorporated dideoxynucleoside monophosphate (ddNMP) at the 3′ end of the primer strand, the incoming ddNTP and two Mg²⁺ ions, for which unambiguous electron density was observed, were built into the model (Fig. 1 ▶). The overall conformation of the protein, DNA, and ddNTP in these three structures is very similar to those in the Klentaq1–DNA–ddCTP complex. The root mean square (r.m.s.) deviation in protein atoms for the ddGTP-, ddATP-, and ddTTP-trapped complexes compared with the ddCTP-trapped complex is only 0.65, 0.57, and 0.57 Å, with r.m.s. deviation in C_α atoms only of 0.34, 0.26, and 0.26 Å, respectively.

Fig. 1.

Simulated annealing omit maps of the incoming ddNTPs and the paired template bases in the ddGTP- (A), ddATP- (B), and ddTTP- (C) trapped ternary complexes. The maps are contoured at a 1.5-σ level. The final refined model is shown in ball-and-stick representation with carbon atoms in yellow, oxygen atoms in red, and nitrogen atoms in blue. The figure was made using the program O (Jones et al. 1991).

Comparison of all the four closed ternary complexes shows that the first single stranded base on the template strand (A_T3) for which electron density was interpretable has widely different conformations. These differences reflect a lack of ordering for this particular nucleotide: indeed A_T3 is solvent-exposed and makes very few contacts with the protein in all four complexes.

Common features of the nascent base pair and its environment

A superposition of the four ternary complex structures using only the protein backbone atoms indeed reveals similarities between complexes in both the structure of the nascent base pair and that of the protein surrounding it (Fig. 2A ▶). Table 1 lists the contacts between protein atoms and base atoms in all four ternary complexes, whereas Figure 3 ▶ presents stereodiagrams of these contacts: as can be seen readily, there are very few base-specific contacts. The overall nascent base pair positioning and configuration is not much affected by the type of base pairing formed between the incoming nucleotide and the corresponding template base: the sugar and phosphate groups of the nascent base pair superimpose well and the bases in all four structures are approximately in the same plane (Fig. 2A ▶).

Fig. 2.

Features of the nascent base pair and the base pair preceding it. (A) Local conformational variability in the binding site surrounding the nascent base pair. Only residues contacting the base atoms of the nascent base pair are shown (see Table 1 for list of contacts). The O and P helices are shown in ribbon representation. The ddTTP-, ddATP-, ddGTP-, and ddCTP-trapped complexes are color-coded green, white, cyan, and lavender, respectively. (B) Stereosurface diagram of the binding site around the nascent base pair in the ddTTP-trapped complex. The protein surface is shown in gray, whereas the nascent base pair surface and that of the base pair immediately preceding is shown in cyan and red, respectively. The stick figures represent the nascent base pair dA : ddTTP with carbon atoms in green, oxygen atoms in red, and nitrogen atoms in blue. (C) The base pair formed by the incorporated ddNMP and the template base. Color coding is the same as in A. The dashed lines represent the hydrogen bonds made between Q754 and R573 and the N3 and O2 atoms in all base pair combinations. Figure was made using the program Ribbons (Carson 1997).

Table 1.

Contacts between the protein and the nascent base pair

		Contacts with the template		Contacts with the incoming ddNTP		Contacts with the template		Contacts with the incoming ddNTP
		dAa DNA/protein	Å	ddTTP DNA/protein	Å	dTa DNA/protein	Å	ddATP DNA/protein	Å
	Major groove	C8: R677 Nη1	3.9	C5: F667 Cδ1	3.8	C5M: M673 Cɛ	3.4	C8: F667 Cδ1	3.5
		G668 Cα	3.7			R677 Cζ	3.6	F667 Cɛ1	3.7
		N7: G668 Cα	3.8			R677 Nη1	3.9	N7: F667 C	3.7
		R677 Nη2	3.8			R677 Nη2	3.6	C5: F667 Cδ1	3.7
		C5: G668 Cα	3.7			C5: M673 Cɛ	3.9	C6: F667 Cβ	4.0
						G668 Cα	3.8
						C4: G668 Cα	3.9
						O4: T664 Oγ1	3.6
dA:ddTTP and dT:ddATP	Minor groove	N9: G668 O	3.8	N1: F667 Cδ1	3.6	N1: G668 Cα	3.4	N9: F667 Cδ1	3.5
		G668 Cα	3.5	F667 Cɛ1	4.0	G668 O	3.6	F667 Cɛ1	3.9
		C4: F667 O	3.9	C2: F667 Cδ1	3.8	C2: F667 O	3.8	C4: F667 Cβ	3.9
		G668 Cα	3.5	O2: Y671 Cɛ1	3.8	G668 Cα	3.5	N3: F667 Cγ	3.8
		N3: Y671 Cβ	3.7			O2:Y671 Cβ	3.3	C2: F667 Cδ1	3.6
		Y671 Cδ1	3.7			Y671 Cγ	3.9
		F667 O	3.4			Y671 Cδ1	3.5
		C2: Y671 Cδ1	3.7			F667 O	3.5
		F667 O	3.7

		Contacts with the template		Contacts with the incoming ddNTP		Contacts with the template		Contacts with the incoming ddNTP
		dGa DNA/protein	Å	ddCTP DNA/protein	Å	dCa DNA/protein	Å	ddGTP DNA/protein	Å
Contact distances in Å are indicated on the right. Note that only contacts with the base groups are listed. Also, the well-conserved water-mediated contact between E615-N750-Q754 and the minor groove of the nascent base pair is not listed.
a Refers to base T4 in Fig. 3 ▶.
(ddNTP) dideoxynucleoside triphosphate; (ddTTP) dideoxythymidine 5′-triphosphate; (ddATP) dideoxyadenosine 5′-triphosphate; (ddCTP) dideoxycytidine 5′-triphosphate; (ddGTP) dideoxyguanosine 5′-triphosphate.
	Major groove	C8: R667 Nη1	3.3	C5: F667 Cδ2	3.8	C5: R677 Cζ	3.8	C8: F667 Cδ1	3.9
		R677 Nη2	3.2	C4: F667 Cδ2	3.7	R677 Nη1	3.7	F667 Cɛ1	3.9
		R677 Cζ	3.4			R677 Nη2	3.6	N7: F667 Cδ1	3.8
		G668 Cα	3.6			N4: T664 Cβ	3.9	R660 Nη1	3.6
		N7: G668 Cα	3.8					R660 Nη2	3.6
		R677 Nη1	3.8					C5: F667 Cδ1	3.5
		R677 N|gn2	2.9					C6: R660 Nη2	3.9
		C5: G668 Cα	3.6					O6: R660 Nη2	3.2
		O6: T664 Cα	3.9
		T664 Oγ1	3.7
dG:ddCTP and dC:ddGTP	Minor groove	N9: G668 Cα	3.3	N1: F667 Cδ2	3.9	N1: G668 Cα	3.7	N9: F667 Cδ1	3.7
		C4: G668 Cα	3.4	C2: F667 Cδ2	3.8	C2: F667 O	3.9	F667 Cɛ1	3.7
		G668 N	3.9			G668 Cα	3.7	C4: F667 Cδ1	3.4
		N3: Y671 Cβ	3.7			O2: F667 O	3.3	N3: F667 Cγ	3.9
		Y671 Cδ2	3.9			F667 C	3.8	F667 Cδ1	3.8
		F667 O	3.5			Y671 Cβ	3.3	C2: F667 Cβ	3.6
		F667 C	3.9			Y671 Cγ	3.9	N2: Y671 Cδ1	3.1
		G668 Cα	3.9			Y671 Cδ1	3.6	Y671 Cɛ1	3.6
		C2: Y671 Cδ2	3.9			G668 Cα	3.9	F667 Cβ	3.9
		F667 O	3.7
		F667 C	3.9
		F667 Cβ	3.8
		N2: F667 Cβ	3.8
		Y671 Cγ	4.0
		Y671 Cδ2	3.1
		Y671 Cɛ2	3.7

Fig. 3.

Stereodiagram of the contacts between the protein and the nascent base pair. Direct contacts between protein and base atoms are shown, as well as the water-mediated contacts between E615, Q754, and N750 and the bases. Direct protein-base contacts are listed in Table 1. Secondary structures are shown in ribbon representation with helices in blue, strands in green, and coils in orange. Contact protein residues are shown in stick representation color-coded in gray for carbons, red for oxygens, and blue for nitrogens. Primer and templates bases are in ball-and-stick representation color-coded in white for carbon, red for oxygens, phosphorus in pink, and blue for nitrogens. The incoming ddNTP is shown in ball-and-stick representation with carbon atoms in green, oxygen atoms in red, phosphorus atoms in pink, and nitrogen atoms in blue. The two metal ions are in gold. The water molecule coordinated by E615, Q754, and N750 in all four complexes is shown in red and labeled W. The incorporated ddNMP is labeled ddA, ddT, ddC, and ddG in the various complexes. The base before that is labeled C_P11. Labeling of the template bases is according to Li et al. (1998b). (A) The ddATP-trapped complex. (B) The ddTTP-trapped complex. (C) The ddCTP-trapped complex. (D) The ddGTP-trapped complex. Figure was made using the program Ribbons (Carson 1997).

Both protein and DNA atoms form a narrow pocket around the nascent base pair (see Fig. 2B ▶ for the ddTTP-trapped complex given as an example). One side of the pocket is formed by the base pair of the incorporated ddNMP and the corresponding template base (surface in red in Fig. 2B ▶). On the other side, the O helix in the fingers domain forms part of the pocket with the aromatic ring of F667 stacking with the base of the incoming ddNTP and the main chain atoms of G668 stacking against the template base (details in Table 1 and Fig. 3 ▶). On the minor groove side of the nascent base pair, part of the side chain of Y671 at the bottom of the O helix lies underneath the template base (Table 1 and Fig. 3 ▶), and the side chain of E615, N750, and Q754 makes sequence-independent interactions with the incoming ddNTP through water-mediated hydrogen bonds with the N3 atom in purine bases or the O2 atom in pyrimidine bases (Fig. 3 ▶). On the major groove side of the nascent base pair, contacts involve residues in the O helix (F667 interacting with the incoming base; G668 interacting with the template base) and with residues in the P helix (R677 interacting with the template base; Table 1 and Fig. 3 ▶). The triphosphate group of the incoming ddNTP is mostly buried in the closed ternary complexes. Interactions with the protein are very similar in all ternary complexes: one side of the triphosphate group forms electrostatic interactions with the conserved side chains of R659 and K663 on the O helix, and the other side interacts with the catalytic core formed by the conserved acidic side chains at the base of the palm domain and two Mg²⁺ ions (result not shown because these interactions in all four ternary Klentaq1 complexes are similar to those observed in the previously reported ddCTP-trapped complex [Li et al. 1998b]).

Overall, the tight packing around the nascent base pair observed in both the ternary ddGTP-trapped T7 polymerase complex and the ternary ddCTP-trapped Klentaq1 complex (Doublié et al. 1998; Li et al. 1998b) are also observed in the ddATP-, ddTTP-, and ddGTP-trapped ternary complexes of Klentaq1 presented here. Although these results were anticipated by us (Li et al. 1998b) and others (Doublié et al. 1998), they are experimentally confirmed here.

Plasticity of the active site binding pocket: Differences in the environment of the nascent base pair

Although the environment of the nascent base pair in all four complexes is similar, differences are apparent and are likely to be significant. Clearly, the enzyme is capable of remodeling its active site structure around the slight chemical variations existing between base pairs. This property, which we term plasticity, may constitute the structural basis for the ability of the enzyme to accommodate four different combinations of base pairing.

On the minor groove side, only very small adjustments to base pairing chemistry are observed, and these only affect residue Y671. In the ddTTP- and ddATP-trapped complexes, Y671 has moved by ∼0.8 Å compared with its position in the ddGTP- and ddCTP-trapped complexes (Fig. 2A ▶). This difference is due to the presence of the N2 amine in the G base either of the incoming nucleotide in the ddGTP-trapped complex or of the template in the ddCTP-trapped complex: this group imposes steric constraints on Y671 requiring its slight displacement (Fig. 2A ▶). However, overall, contacts between the minor groove atoms and the protein side chains are similar for all the four ternary complexes. All the minor groove atoms in the nascent base pair are in contact with protein side chains (Table 1), and the geometry of the minor groove pocket is strictly defined by the side chain of Y671 and the E615-N750-Q754–coordinated water molecule (Figs. 2A, 3 ▶ ▶).

In contrast, on the major groove side of the nascent base pair, the geometry of the base pair is relatively different in the four complexes and the side chains of the protein residues undergo slight rearrangements to accommodate the structural variations between base pairs. For the dA : ddTTP and dG : ddCTP base pairs, the C5, N7, and C8 atoms common to dA and dG, and the C5 and C4 atoms common to ddTTP and ddCTP are in contact with the protein. Among the atoms that make each base pair distinct, only the O6 atom of the dG base in the ddCTP-trapped complex is within 3.7 Å from a protein residue, T664. Other atoms (N6 in dA, O4 and the C5 methyl group in ddTTP, and N4 in ddCTP) are not in contact with protein side chains. Overall, the major groove side of the pocket in the ddTTP- and ddCTP-trapped complexes is slightly more open than in the ddGTP- or ddATP-trapped complexes.

For the ddGTP- and ddATP-trapped complexes, some side chains in the protein adopt different conformations to fit the geometry of the major groove of the dC : ddGTP and dT : ddATP base pairs. The side chain of R677 contacts the C5 atom of the dC base in the ddGTP-trapped complex (the same position is occupied by the N7 atoms in the ddTTP- and ddCTP-trapped complexes), and it moves outward by ∼1 Å in the ddATP-trapped complex, to accommodate the methyl group of the dT base on the template. The side chain of T664 contacts the O4 group of the dT base in the ddATP-trapped complex and the N4 group of the dC base in the ddGTP-trapped complex. M673 makes contact with the C5 methyl of dT in the ddATP-trapped complex: this is the only complex where this residue is involved in contacts with the incoming nucleotide. No specific sequence interactions between the major groove side of the ddATP and the protein are found. However, there are base specific hydrogen-bonding interactions between the ddGTP and the protein (described below).

Among the four closed complexes, the ddGTP-trapped complex differs from the other three in that the side chain of R660 interacts with the base of the incoming ddGTP and the side chain of R587 forms hydrogen bonds with the N7 atom of the incorporated ddGMP. As a result of the hydrogen-bonding interactions between these two side chains and the major groove of the DNA, the tip of the fingers domain (the loop between helices N and O and part of helix O) in the ddGTP-trapped complex is moved closer to the active site than those in the other three complexes. The r.m.s. deviation in C_α atoms in this region between the ddGTP-trapped complex and the ddCTP-trapped complex is 1.2 Å, whereas that between the ddATP- or ddTTP-trapped complex and the ddCTP-trapped complex is ∼0.4 Å.

Taq polymerase incorporates ddGTP at a rate ∼10 times higher than that observed for ddATP, ddTTP, or ddCTP (Brandis et al. 1996). The specific contacts made by R660 in the ddGTP-trapped complex were shown to account for this effect (Li et al. 1999). Whether these contacts also are made in the dGTP-bound state of the enzyme or only occur during incorporation of ddGTP is not known. Because the additional hydrogen-bonding interactions made by R660 and R587 in the ddGTP-trapped complex are to the base of the nucleotide, one would expect the conformational changes affecting these residues to occur also during incorporation of dGTP. However, previous studies have shown that the rate of dGTP incorporation by Taq polymerase is similar to that of dATP, dTTP, or dCTP (Brandis et al. 1996). These results suggest that the conformational changes affecting R660 and R587 may be slow processes that occur only during the much slower incorporation of dideoxynucleotides (Brandis et al. 1996).

Minor groove recognition in the base pair preceding the nascent base pair

The base pair preceding the nascent base pair is formed by the incorporated ddNMP and the corresponding template base. Its major groove (where many of the differences between base pairs lie) faces the solvent, and therefore all differences in protein structure between complexes are observed on the side of the minor groove; these are very small, mostly affecting the local conformation of the side chains of residues Q754 and R573 (Fig. 2C ▶) that must adjust to the steric constraints imposed by the N2 amine group of the G base in either the dG : ddCMP or the dC : ddGMP base pair.

In all four complexes, the N3 atom of purine bases or O2 atom of pyrimidine bases in the ddNMP or those in the corresponding template base are superimposable (Fig. 2C ▶). These atoms make hydrogen-bonding interactions with the side chains of Q754 and R573. Residues Q754 and R573 are conserved among all Pol I family enzymes, and similar hydrogen-bonding interactions between homologous residues and bases have been observed in the T7 DNA polymerase (Doublié et al. 1998) and the Bacillus stearothermophilus DNA polymerase systems (Kiefer et al. 1998). Because only Watson-Crick base pairing can result in a pseudo twofold symmetrical arrangement of the hydrogen-bonding acceptors N3 of purines and O2 of pyrimidines, recognition of these atoms by Q754 and R573 may provide a means of detecting a properly matched base pair. This hypothesis, however, remains to be tested experimentally.

Conclusion

In this report, we present a unique collection of ternary complex structures, which provide insights into recognition of base pairs by DNA polymerase I enzymes. These structures unravel an ability of the enzyme to model itself around the various base pairing geometries. Plasticity also may play a role in fidelity during template-directed DNA synthesis.

Kinetic studies have shown that E. coli Klenow does not show discrimination against a mismatched nucleotide during the ground-state dNTP-binding event (Johnson 1993). However, the rate-limiting conformational change before chemistry is slowed drastically during mismatch incorporation (Johnson 1993). These results suggest that the rate-limiting conformational change, corresponding to the open to closed transition in the fingers domain (Doublié et al. 1998; Li et al. 1998b), may account for nucleotide selectivity during nucleotide incorporation. Because we show here experimentally that there are only very few sequence-specific contacts between the protein and the incoming nucleotide, we infer that selectivity must originate from the steric constraints imposed on the nascent base pair by the enzyme in the closed form: indeed, in all four complexes, the narrow pocket formed by the protein around the nascent base pair is best suited to accommodate a correctly formed Watson-Crick base pair.

In support to the steric selection mechanism, studies using the E. coli Klenow fragment have shown that mutating residue Y766 to Ser increases the incorporation of a mismatch up to 44-fold (Carroll et al. 1991). Y766 in E. coli Pol I is the equivalent residue of Y671 in Klentaq1. The aromatic ring of this residue forms part of the active side pocket at the minor groove side of the nascent base pair in all four ternary complexes (Fig. 3 ▶). Replacement of the bulky ring with a much smaller serine side chain would reduce the steric constraints placed on the nascent base pair, thereby possibly allowing a mismatch to fit in the active site.

The proposed steric selection mechanism is supported by the studies of Moran et al. (1997). These authors show that a shape analog of thymidine completely lacking hydrogen-bonding ability (base F), can be incorporated efficiently into the primer strand at positions opposite to an adenine base by the Klenow fragment to form an A : F base pair. The solution structure of a duplex DNA containing the A : F base pair shows that it has the same shape as an A : T base pair, and it causes little distortion to the duplex DNA structure (Guckian et al. 1998). These results suggest that the shape of the base pair, not the hydrogen-bonding interactions between the bases, determines the nucleotide selectivity during the polymerase reaction (Goodman 1997).

The discussion on nucleotide selectivity presented here is based on the assumption that mismatched base pairs are structurally distinguishable from the Watson-Crick base pairs. The shape of a mismatch formed by two purines or two pyrimidines is clearly different. However, for purine : pyrimidine mismatches, this is true only if the shape of these base pairs (wobble base pairs) is determined by the hydrogen-bonding interactions between the bases. In fact, in the case of a purine : pyrimidine mismatch, two opposite forces may come into play: hydrogen-bonding interactions between the bases favor the formation of a wobble base pair, but the steric constraints from the polymerase favor the formation of a base pair that has the shape of a Watson-Crick base pair. Hydrogen-bonding interactions between bases may have dominant effects so that a wobble base pair is formed; thus, a stable tight binding complex cannot form, and mismatch incorporation does not occur. However, depending on the sequence contents of the template, the concentration of the various nucleotide substrates, or the type of metal ions, the steric constraints imposed by the protein may force the mismatch into a shape closer to that of a Watson-Crick base pair able to fit into the active site of the enzyme in its closed form. In fact, the intrinsic plasticity of the active site (that the ternary complex structures presented here reveal) suggests that intermediate structures between a Watson-Crick base pair and a wobble base pair may be accommodated by the enzyme and still form closed, albeit less stable, complexes with the polymerase. We then conclude that, although designed to be highly accurate in the processing of DNA, DNA polymerase I enzymes also may have evolved a means to make occasional mistakes.

Materials and methods

The experimental design, crystallization, and structure determination of the three complexes of Klentaq with the DNA and ddATP, ddTTP, and ddGTP were described previously (Li et al. 1999). The statistics of the data collection and structure refinement also were shown in the same report (see Table 1 in Li et al. 1999).