Limitation of existing methods of studying catalysis
Enzymes and associated catalytic reactions have been studied in the pre-steady and steady states to yield catalytic requirements, intermediate steps and rates at high temporal resolutions. DNA synthesis reactions are no exception. DNA polymerases catalyze phosphoryltransfer reactions that incorporate dNTPs (A, G, T and C) according to a template sequence one at a time into DNA primer (Fig. 1a; Additional file 1: Movie S1). The reaction is of SN2 type and is pH dependent. All DNA polymerases depend on the metal ions Mg2+ or Mn2+ for catalysis, but differ dramatically in their catalytic rates. Many DNA polymerases undergo dNTP-dependent large conformational changes, while some translesion polymerases do not [12, 26–30]. Extensive kinetic and FRET studies have shown that the large conformational changes in DNA polymerases are faster than the chemical reaction itself, and therefore the phosphoryltransfer reaction is the most critical step in all DNA polymerases [9, 31–33]. After decades of investigation, it remains uncertain how a 3′-OH nucleophile is activated and whether the rate-limiting step of DNA synthesis is chemical or conformational [12, 32, 34–37].
For high spatial resolution, a large number of DNA polymerases have been crystallized in complex with a DNA template and primer pair and incoming dNTP. Catalysis was invariably circumvented by removal of the nucleophilic 3′-OH (replaced by 3′-deoxyribose) at the primer end, substitution of Mg2+ or Mn2+ by Ca2+, which does not support the catalysis, or substitution of dNTPs by non-reactive ddNTP or dNMPNPP analogs [14–19, 38]. Despite the diverse tertiary structures, catalytic rates, and degrees of fidelity in DNA polymerases, the core of the active site contains two Mg2+ or Mn2+, which are coordinated by the conserved carboxylates, DNA and dNTP substrate in the same configuration [19, 38]. How catalysis takes place was not visualized except for by QM/MM (hybrid quantum mechanics/molecular mechanics) modeling [20–23], which never fails to show that products can form by rearranging protons, electrons and atoms.
Kinetic studies of DNA synthesis have also observed two different binding constants of Mg2+ [37]. But if more than one metal ion has a similar binding affinity, such metal-ion titration assays would lead to an under estimation. Because kinetic titration results agree with the number of metal ions found in the active site by X-ray crystallography [14–16, 26], two-metal-ion catalysis has become the widely accepted mechanism for DNA synthesis reactions. In the two-metal-ion catalysis model, one metal ion (B site) is associated with incoming dNTP and the second metal ion (A site) bridges the 3′-OH nucleophile and the incoming dNTP and is suggested to help deprotonating 3′-OH for the nucleophilic attack (Additional file 1: Movie S1). If incorrect or damaged, an incoming dNTP would be rejected before the A-site Mg2+ is recruited. Therefore, binding of two Mg2+ ions, which are notorious for the stringent coordination requirement, helps polymerases to achieve exceedingly high fidelity [39, 40]. As initially proposed by Steitz and Steitz [41], the DNA synthesis reaction is presumed to be promoted by the two Mg2+ ions, which align the substrates and facilitate the acid–base catalysis and the pentacovalent intermediate formation (Additional file 1: Movie S1).
In crystallo catalysis of DNA synthesis
Because DNA synthesis is pH and metal ion dependent, by reducing pH to 6.0 and using the non-catalytic metal-ion cofactor Ca2+, crystals of native DNA Pol η in complex with DNA substrate and correct incoming dNTP can be grown over a couple of weeks without reaction taking place [24]. After fully grown, crystals were transferred to stabilization buffer to remove free reaction components and raise the pH to 7.0, which is permissible for the phosphoryltransfer reaction. To initiate DNA synthesis, crystals were exposed to Mg2+ or Mn2+, and after a certain reaction time they were cryo-cooled in liquid nitrogen to stop the reaction at the set time point. Ensuing analyses by X-ray diffraction have yielded unprecedented details following the reaction process of DNA synthesis [24]. Binding of the A and B site Mg2+ is necessary to fully align the 3′-OH of DNA primer and incoming dNTP. But only after a 40 s delay after two metal ion binding, does the chemical reaction take place [24].
Interestingly, in crystallo DNA synthesis has also been achieved with DNA polymerase β. Crystals of DNA polymerase β complexed with DNA substrate (binary complexes) have been grown; however, unlike our Pol η studies, the incoming dNTP and Ca2+ were soaked in later [42]. Similar to the studies of Pol η, catalysis was initiated by exposing Pol β crystals to Mg2+ or Mn2+ and terminated by cryo-cooling in liquid nitrogen. This approach has been applied successfully to study how Pol β incorporates an incorrect incoming nucleotide and oxidatively damaged 8-oxo-dG [42, 43]. Because the reaction rate of Pol β is much faster than Pol η, at the first time point of the data acquisition half of the substrate is already turned into product. Substrate alignment by two metal ions and the delayed chemical reaction were not observed with Pol β [42, 43].
A third metal ion is essential for DNA synthesis
In both in crystallo catalysis by Pol η and β, a third divalent cation coordinated by reaction products is observed (Fig. 1b–d). In the Pol β case, the third metal ion appears at the first time point (20–30 s) when approximately half of the substrate was converted to products. With Pol η, whose reaction rate is much slower, the third Mg2+ was detected ~60 s after the initial appearance of products, when the products were accumulated to 40 % [24]. Because the third Mg2+ is coordinated by the DNA product and pyrophosphate leaving group and its binding is incompatible with the substrate state, it is dubbed “the product metal ion” [42, 43].
We were puzzled about the role of the third metal ion and wondered whether it is involved in catalysis or merely stabilizes the products and acts as a general acid to facilitate protonation of pyrophosphate [24]. Because of the low electron number of Mg2+, it is not possible to detect Mg2+ by X-ray diffraction if its occupancy is lower than 30 %. Therefore, even if the third Mg2+ is required for the catalysis and product formation, it is not detected until the amount of product reaches more than 30 %.
To detect the third metal ion at a low occupancy, we switched the reaction buffer from Mg2+ to electron-rich Mn2+. To our delight, Mn2+ not only allowed detection of the third metal ion at low occupancy, but the affinities for Mn2+ at the two canonical (A and B) and the third (C) sites are significantly different [25]. When in crystallo reaction took place at 10 mM Mn2+, the third Mn2+ appears at the same time as the reaction products and the two are perfectly correlated in appearance, time and occupancy. When the reaction was conducted in 1 mM Mn2+, however, the A and B sites were readily occupied, but the C-site was devoid of Mn2+. As a result, no products could be detected! Our in crystallo metal ion titration reveals the canonical A and B sites both have high-affinity for Mg2+ and Mn2+, and the C-site is of low affinity and thus determines the overall metal-ion concentration requirement for the DNA synthesis reaction [25, 37].
Binding of the third metal ion is rate limiting
A protein sidechain, R61 of Pol η, forms bifurcated salt bridges with the α and β phosphates of the incoming dNTP in the reactant state and overlaps with the C-site metal ion in the polymerase-product complexes if R61 does not change its rotamer conformation [24]. One may imagine that removal of R61 would allow easier and faster binding of the third metal ion and lead to a faster reaction rate. When R61A mutation is introduced, however, the mutant Pol η catalyzes DNA synthesis at a slower rate than WT polymerase [19, 44, 45]. When the R61A mutant Pol η was examined in crystallo, we found that binding of the A and B site metal ions occurred within 40 s as for WT, but without R61 the incoming dNTP is misaligned relative to the 3′-OH by 0.3 Å. Consequently binding of the third metal ion was delayed by 120 s, thus leading to the slower catalytic rate than WT [25].
All DNA polymerases exhibited reduced catalytic rates when incoming dNTP is substituted with sulfur in the pro-Sp position (Sp-dNTPαS) [8, 34–36]. Before our discovery of the third metal ion, the Sp-sulfur was thought not to be involved in metal ion binding, and the degree of rate reduction has been interpreted as perturbing either chemistry or conformational change necessary for catalysis [46]. But the pro-Sp and the α,β bridging oxygen atoms of dNTP are the only two potential non-water ligands of the third metal ion (Fig. 1). We predicted that sulfur substitution of Sp-dNTPαS therefore would retard the third metal ion binding and reduce the reaction rate.
Experimentally we find that the concentrations of Mg2+ and Mn2+ needed for incorporating Sp-dNTP αS by Pol η in solution are increased by ten and threefold, respectively [25]. This is not surprising because the amount of Mg2+ or Mn2+ needed for the DNA synthesis reaction is determined by the C-site metal ion [25]. The most dramatic changes in the in crystallo reaction are slow product formation and absence of detectable third metal ion even in 20 mM Mg2+ or Mn2+. In addition, although not directly involved in the A and B site metal ion binding, the Sp-sulfur perturbs the occupancy and location of Mg2+ in the A site [25] probably due to an altered electrostatic environment. Because the reaction-rate reduction caused by Sp-dNTP αS is less than fourfold [35, 36], it was concluded that the rate-limiting step in DNA synthesis is conformational changes. But the high concentrations of Mg2+ (up to 12.5 mM) used in these assays, however, masked the severe defects of the C-site metal ion binding. If 1 mM of Mg2+ had been used, the rate reduction would have been much greater, and thus leading to the opposite conclusion that chemistry is the rate-limiting step!
Our study reveals that binding of the C-site metal ion is essential and that its occurrence determines when the chemical reaction takes place. Binding of the two canonical metal ions at the A and B sites is a prerequisite for capturing the third metal ion but it is not rate limiting. Because delays of the C-site metal ion binding reduce the reaction rate as observed with R61A Pol η and Sp-dNTPαS substitution, we conclude that binding of the third divalent cation is the rate-limiting factor in the DNA synthesis reaction.
Solving the conundrum of when and how the third metal ion binds
The requirement of the third metal ion for the DNA synthesis reaction and its absence in the enzyme-substrate state present a conundrum—when and how the third metal ion binds. With nothing else to turn to, we hypothesized that thermal energy and thermal motion of the well-aligned reactants might provide a transient entrance for the third metal ion. To test the hypothesis, we divided the DNA synthesis reaction into two stages, A- and B-site metal ion binding along with alignment of DNA and dNTP substrate as the first, and capture of the third metal ion and product formation as the second. The first stage was carried out at 1 mM Mn2+. The second stage was conducted in the presence of 5 mM Mn2+ at various temperatures. Variation from 4 to 37 °C did not alter the diffusion rate of Mn2+ as evident in its unchanged binding at the canonical A site, but the increased thermal motions significantly enhanced the third metal ion binding and product formation [25].
The coordination of the third metal ion by two oxygen atoms of an incoming dNTP is non-ideal in distance and geometry compared to the preferred octahedral geometry of six inner-shell ligands of Mg2+ ion. But its coordination in the product state, when the phosphodiester bond between the α and β phosphates is broken, is nearly perfect octahedral. We suspect that the free- energy gain from binding of the third Mg2+ ion (freeing two inner-shell water ligands and association with dNTP) clears the barrier to the transition state. The stringent preference for octahedral coordination by Mg2+ ion may also drive the phosphotransfer reaction from breaking the existing bond in dNTP to forming a new bond between the 3′-OH and the α phosphate, which is in the reverse direction of the standard textbook version that is starting from the nucleophilic attack to the bond breakage (Fig. 2).