Capture of A Third Mg²⁺ is Essential for Catalyzing DNA Synthesis

Abstract

It is generally assumed that an enzyme-substrate (ES) complex contains all components necessary for catalysis, and conversion to products occurs by rearrangement of atoms, protons and electrons. However, we find that DNA synthesis does not occur in a fully assembled DNA polymerase-DNA-dNTP complex with two canonical metal ions bound. Using time-resolved X-ray crystallography, we show that the phosphoryltransfer reaction takes place only after the ES complex captures a third divalent cation that is not coordinated by the enzyme. Interestingly, binding of the third cation is incompatible with the basal ES complex and requires thermal activation of the ES for entry. We suggest that the third cation provides the ultimate boost over the energy barrier to catalysis of DNA synthesis.

Enzymes increase the rate of chemical reaction, which is thought, to occur by reducing the activation energy required to reach the transition state (1–3) (Fig. 1A). Due to their transient and unstable nature, authentic transition states have not been visualized but are assumed to have the same chemical components as the substrate state. DNA polymerases, which catalyze a phosphoryltransfer reaction that incorporates dNTPs into DNA, are known to require two Mg²⁺ ions (4–8) (Fig. 1B). Despite extensive kinetic studies using the stopped flow technique and the dNTP analog dNTPαS, it remains controversial whether a conformational transition prior to catalysis (9–14) or the chemistry itself (15–16) is the rate-limiting step in DNA synthesis.

Figure 1 |. No DNA synthesis without the third metal ion.

A, Reaction coordinate of enzyme catalysis. B, The assumed transition state of the two-metal-ion catalysis. C, The structure of Pol η catalyzing DNA synthesis in crystallo (PDB: 4ECV, 18). The C-site Mg²⁺ is coordinated by the products (60%, blue) but clashes with the substrate dATP and R61 side chain (40%, yellow). D, The structure of Pol η incubated with substrates and 1 mM Mn²⁺ for 1800s. No C-site metal ion or reaction products were detected. The F_o-F_c map with the active center omitted contoured at 3.5σ level (blue meshes) is superimposed.

We have recently visualized phosphodiester bond formation catalyzed by human DNA Pol η in crystallo (17). Consistent with the two-metal-ion mechanism (6–8), binding of Mg²⁺ ions in the A and B sites occurs within 40 s leading to alignment of the 3´-OH of the primer end with the α-phosphate of dNTP (Fig. 1C) (17). After another 40 s, product starts to appear without discernible conformational change of the enzyme or substrates. However, we observed a third Mg²⁺ ion appearing in a third ‘C’ site after product formation (Fig. 1C) (17). An equivalent third metal ion coordinated by the reaction products and four water molecules has also been observed in the in crystallo catalysis by DNA Pol β (18–20). Due to steric clashes with dNTP (Fig. 1C), the third metal ion cannot bind in Pol η enzyme-substrate complexes. Because of low occupancy in the C site and weak diffraction of Mg²⁺ ions, it has been unclear when the third Mg²⁺ appears and whether it is involved in the phosphoryltransfer reaction.

To determine the reaction coordinates of Pol η and the role of the third metal ion, we replaced Mg²⁺ with Mn²⁺, which supports DNA synthesis (21) and is readily detected by X-ray diffraction even at low occupancy. Crystals of native Pol η (1–432 aa) complexed with DNA, dATP and Ca²⁺ were grown at pH 6.0 in a non-reactive ground state (17). After exposure to a pH 7.0 reaction buffer containing 1 mM Mn²⁺ for 90 to 1800 s, crystals were flash frozen in liquid N₂ to stop the reaction, and 1.5–1.7 Å X-ray structures were determined at five reaction time points (table S1 and Materials and Methods). All five structures were practically identical except for the gradual replacement of Ca²⁺ by Mn²⁺ in the B site (fig. S1). By 600 s when ~90% of the A and B sites were occupied by Mn²⁺, the 3´-OH of the DNA primer was aligned with the α-phosphorus of the dATP, and the structure was identical to that of crystals soaked in 1 mM Mg²⁺ for 40 s (fig. S2, A-B). Similar to the reaction in Mg²⁺ (17), a water molecule (WatN) hydrogen bonded to the nucleophilic 3´-OH appeared at 90 s, and its occupancy increased with time in correlation with binding of the A-site Mn²⁺ (fig. S2, C-D). However, in 1 mM Mn²⁺, the Pol η-DNA complex remained in the substrate state with no product and no C-site Mn²⁺ ion for up to 1800s (Fig. 1D).

We assayed the metal ion (Me²⁺) requirements for Pol η catalysis in solution and found that 0.6 mM Mg²⁺ or 2.7 mM Mn²⁺ is needed to attain the half-maximal reaction rate (Fig. 2A, table S2). We then examined the Mn²⁺ affinity of each binding site in crystallo. While increasing the Mn²⁺ concentration (0.5 to 15 mM) accelerated the rate of metal-ion binding in all three sites (Fig. 2B, table S3), the apparent K_d’s of the A and B sites were below 0.5 mM. The K_d for the C site, however, was 3.2 mM, close to the 2.7 mM measured in solution (Fig. 2A).

Figure 2 |. Coupled appearance of the third Mn²⁺ and reaction products.

A, Mg²⁺ (purple) and Mn²⁺ (green) dependence of Pol η catalysis in solution. B, Titration of the A-, B-, and C-site Mn²⁺ binding in crystallo. The 600s data were fitted to equilibrium binding modes to yield the K_d’s. C, Structures of Pol η in crystallo catalysis with 10 mM Mn²⁺. The F_o-F_c omit map for the new bond, the C-site Mn²⁺ (blue) and the WatN (pink) were contoured at 3σ and superimposed onto each structure. D, Correlation between the new bond formation and the C-site Mn²⁺ binding.

When in crystallo reactions occurred in 10 mM Mn²⁺, catalysis proceeded as in 1 mM Mg²⁺ (17), except that the A-site Mn²⁺ did not dissociate upon product formation as does Mg²⁺ (fig. S3, A-B) and slightly less product accumulated with Mn²⁺ than Mg²⁺. However, unlike the reaction in Mg²⁺, the C-site Mn²⁺ appeared simultaneously with the reaction product, 30 s after binding of the two canonical metal ions (Fig. 2C). Electron density for the new phosphodiester bond and the C-site Mn²⁺, whose chemical nature was confirmed by its anomalous diffraction and characteristic octahedral coordination geometry (fig. S3C), had one-to-one correlation at every time point and Mn²⁺ concentration (Fig. 2D). In Mg²⁺ by contrast, with 15% product formed at 80 s the C-site Mg²⁺ was at too low occupancy to be observed and was not detected until 140 s when product had accumulated to 40% (fig. S3A) (17). Previous stopped-flow studies indicate that one of the metal ion-binding sites has much lower affinity for Mg²⁺ and thus limits DNA synthesis (16). Our in crystallo titrations unequivocally show that the low-affinity binding site is neither A nor B, but the C site, which determines the concentration of Mg²⁺ or Mn²⁺ necessary for the DNA synthesis reaction.

The C-site Me²⁺ is coordinated by four water molecules and two oxygen atoms, one each from the product DNA and pyrophosphate, which correspond to the α pro-S_p oxygen and the α, β bridging oxygen of dNTP (Fig. 1C). Sulfur substitution of the pro-S_p oxygen (S_p-dNTPαS) has been widely used to dissect the reaction kinetics of DNA synthesis (11–14, 22) because the pro-S_p atom is not directly involved in A- or B-site Me²⁺ coordination. The reduction of the reaction rate by S_p-dNTPαS has been interpreted to be “conformational” (if < 4 fold reduction) or to affect the chemistry itself (if 4–11 fold) (11–14, 23).

As a ligand of the 3^rd Me²⁺, the sulfur in S_p-dATPαS was tolerated by Pol η (table S2) but required much higher [Mg²⁺] (15 mM) and [Mn²⁺] (9 mM) than dATP for catalysis to occur (Fig. 2A, Fig. 3A). Unexpectedly, in crystallo S_p-dATPαS slowed Mg²⁺ and Mn²⁺ binding at the A site. After a lengthened delay product started to form, but the C site remained empty (Fig. 3B, fig. S4A-C). We suspect that the third Mn²⁺ still assisted product formation, but the association was too transient to be observed. In addition, while A-site Mg²⁺ occupancy was reduced in the presence of dATPαS, an alternative A’ site appeared 2.6 Å away (fig. S4, D-E). These data suggest that the reduced reaction rate with S_p-dATPαS cannot be attributed to “conformational” effects (11–14) but involves impaired A- and C-site Mg²⁺ binding and altered reaction chemistry.

Figure 3 |. Changing the C-site environment alters Pol η catalysis.

A, Mg²⁺ (purple) and Mn²⁺ (green) dependence of Pol η incorporating dATPαS in solution. B, In crytallo incorporation of dATPαS by Pol η with 20 mM Mn²⁺ at 600s showed product formation (50%) but no C-site Mn²⁺. The 2F_o-F_c map contoured at 2σ level (blue meshes) is superimposed. C, Time delay in product formation by WT (magenta) and R61A (cyan) Pol η in crystallo. D, Deviation of dATP in the ES of R61A Pol η (cyan with 2F_o-F_c map contoured at 1.5σ) from WT Pol η (magenta).

To bind the third Me²⁺, the R61 side chain, which forms salt bridges with the dNTP (17), moves to vacate the C site (Fig. 1C, 3B). Mutation of R61 to Ala reduces the k_cat by two thirds (table S2) (24–25), but the metal-ion requirement and the general reaction process in crystallo were indistinguishable from WT Pol η (fig. S5E-F). However, the delay between binding of two Mg²⁺ ions and product formation was lengthened from WT’s 40 s to R61A’s 160 s (Fig. 3C). This delay likely stems from a slight shift of dATP away from the active site and a 0.3 Å increase in separation between the 3´-OH and α-phosphate (Fig. 3D). The void left by the R61A mutation was occupied by water molecules (25) and not by the abundant K⁺ or Rb⁺ (identifiable by anomalous diffraction) in the reaction buffer (fig. S5). The subtle misalignment of the substrate, which was repeatedly observed with R61A and R61M mutant Pol η (25) and with dATPαS, led to a prolonged delay before C-site Me²⁺ binding and product formation (fig. S4, B-C).

Notably a +1 charged side chain at the position equivalent to R61 is found in all A-, B-, and Y-family DNA polymerases and reverse transcriptases despite diverse structures of finger domains surrounding the C-site (fig. S6). Among C- and X-family DNA polymerases, there is no R61 equivalent, but the third metal ion has been observed for the X-family Pol β (18–20). The finger domains, which carry the +1 charged residue, distinguish right from wrong correct from incorrect incoming nucleotides by enclosing only a correct dNTP (26). A closed finger appears to be a prerequisite for C-site metal ion binding and catalysis. The varied environment surrounding the C site may thus be exploited for drug design to increase specificity and reduce toxicity of broadly used nucleoside and nucleotide analogs targeting DNA polymerases in antiviral and anti-cancer therapeutics (27–28).

Because the C site does not exist in the Pol η-substrate complex but is required for product formation, we hypothesized that thermal motion of the well-aligned reactants in the ES may create an opening for the 3^rd metal ion. If so, elevated temperature would promote C-site metal ion binding and thus catalysis. To test this hypothesis, we designed a two-step in crystallo reaction. The Pol η crystals were first soaked in 1 mM Mn²⁺ to saturate the A and B sites, and then exposed to 5 mM Mn²⁺ at 4°C to 37°C for 60 s for catalysis to occur (Fig. 4A). The diffusion rate of Mn²⁺ in crystallo was unaffected by temperature as demonstrated by Mn²⁺ binding at the A-site (Fig. 4B). But in the two-step reaction, no C-site Mn²⁺ or product was detected at 4°C (Fig. 4C). At 14°C, low levels of the third Mn²⁺ ion and products were observed, and their amounts doubled at 30°C. The temperature dependence of C-site and product formation corroborates that binding of the third metal ion is rate limiting in the DNA synthesis reaction.

Figure 4 |. Thermal energy-dependent C-site formation and its metal-ion selectivity.

A, Schematic diagram of the two-step in crystallo reactions that probe the C-site formation and ion selectivity. B, Binding of the A-site Mn²⁺ was unaffected by varying temperature from 4°C to 37°C. C, Binding of the C-site Mn²⁺ and the product formation increased with the temperature from 4°C to 37°C. D, Rates of product formation with five metal ions tested in the second step.

To determine the metal ion selectivity at the C site, we varied Me²⁺ in the second step of the two-step reaction (Fig. 4A). Catalysis occurred most efficiently with Mg²⁺, followed by Mn²⁺ and Cd²⁺ (Fig. 4D, fig. S7). Ca²⁺ and Zn²⁺ seemingly also led to product formation at ~40% efficiency of Mg²⁺. However, the C-site coordination geometry with all five Me²⁺ tested appeared identical to that of Mg²⁺ and Mn²⁺ (fig. S7) despite different coordination distances of Ca²⁺ (2.3–2.5Å) and Mg²⁺ or Mn²⁺ (2.1Å). It is thus likely that Ca²⁺ and Zn²⁺ replaced A or B site Mn²⁺ in some Pol η molecules, and the freed Mn²⁺ ions may occupy the C site in other Pol η molecules to support the catalysis. The low affinity and strong preference for Mg²⁺ at the C site, which cannot be explained by its coordination ligands, suggest a catalytic role for the third metal ion in DNA synthesis.

Based on the requirement for three metal ions in DNA synthesis, we suggest a revision of the catalytic mechanism (Fig. 5). DNA synthesis begins with binding of dNTP along with the B-site Mg²⁺ and formation of a ground-state Pol η-DNA-dNTP-Mg²⁺ complex (GS). Watson-Crick pairing between the template and dNTP favors A-site Mg²⁺ binding. The two Mg²⁺ ions and the R61 side chain neutralize and align dNTP with DNA in the reaction-ready state (ES), where the juxtaposed and polarized substrates recruit WatN (fig. S2) (17). However, neither deprotonation nor chemistry takes place without the C-site Mg²⁺. Thermal motion may transiently bring the perfectly aligned reactants closer to each other by fractions of an Ångstrom and create an entry for the third Mg²⁺. Close approach of the reactants may also increase negative charge around the α-phosphate and favor replacement of the +1 charged R61 by the C-site Mg²⁺. We hypothesize that the energy barrier to the transition state is overcome by binding of the third Mg²⁺. The stringent octahedral coordination geometry of Mg²⁺ implies that the C-site Mg²⁺ may help to break the α-β phosphodiester bond (Fig. 5) in addition to protonating the pyrophosphate leaving group (17). Product formation is coupled to disappearance of WatN (fig. S3D), which likely deprotonates the 3´-OH, and to release of the A-site Mg²⁺, which prevents the reverse reaction (fig. S3, A-B).

Figure 5 |. Mechanism of Pol η catalysis.

Pol η binds DNA and an incoming dNTP along with the B-site Mg²⁺ to form the ground state (GS). Binding of the A-site Mg²⁺ leads to the reaction-ready ES state, in which the 3’-OH is aligned with dNTP and WatN recruited. Thermal motions of the reactants create the C site, leading to the third metal-ion binding and the transition state (TS) formation. The C-site metal ion promotes the phosphoryltransfer from the leaving group to the nucleophilic 3´-OH, thus overcoming the energy barrier to the product state (PS).

It has long been assumed that enzymes stabilize transition states and reduce the energy barrier to product formation (Fig. 1A), but de novo design of enzymes based on this assumption has not been successful (29–32). Notwithstanding its crucial role in catalysis, the C-site metal ion of polymerases has evaded detection by biochemical and structural studies of DNA polymerases for decades. Identification of the essential third metal ion in the Pol η catalysis leads us to anticipate that acquisition of transient metal-ion cofactors in transition states may be a general feature that enables enzyme catalysis.