ESR spectroscopy identifies inhibitory Cu²⁺ sites in a DNA-modifying enzyme to reveal determinants of catalytic specificity

Abstract

The relationship between DNA sequence recognition and catalytic specificity in a DNA-modifying enzyme was explored using paramagnetic Cu²⁺ ions as probes for ESR spectroscopic and biochemical studies. Electron spin echo envelope modulation spectroscopy establishes that Cu²⁺ coordinates to histidine residues in the EcoRI endonuclease homodimer bound to its specific DNA recognition site. The coordinated His residues were identified by a unique use of Cu²⁺-ion based long-range distance constraints. Double electron-electron resonance data yield Cu²⁺-Cu²⁺ and Cu²⁺-nitroxide distances that are uniquely consistent with one Cu²⁺ bound to His114 in each subunit. Isothermal titration calorimetry confirms that two Cu²⁺ ions bind per complex. Unexpectedly, Mg²⁺-catalyzed DNA cleavage by EcoRI is profoundly inhibited by Cu²⁺ binding at these hitherto unknown sites, 13 Å away from the Mg²⁺ positions in the catalytic centers. Molecular dynamics simulations suggest a model for inhibition of catalysis, whereby the Cu²⁺ ions alter critical protein-DNA interactions and water molecule positions in the catalytic sites. In the absence of Cu²⁺, the Mg²⁺-dependence of EcoRI catalysis shows positive cooperativity, which would enhance EcoRI inactivation of foreign DNA by irreparable double-strand cuts, in preference to readily repaired single-strand nicks. Nonlinear Poisson-Boltzmann calculations suggest that this cooperativity arises because the binding of Mg²⁺ in one catalytic site makes the surface electrostatic potential in the distal catalytic site more negative, thus enhancing binding of the second Mg²⁺. Taken together, our results shed light on the structural and electrostatic factors that affect site-specific catalysis by this class of endonucleases.

The biochemical basis of specificity in the interaction of proteins with DNA sites is a major problem of modern molecular genetics. Studies of many protein-DNA complexes by crystallography have elucidated the intermolecular recognition contacts (1), but it is clear that point-to-point contacts cannot fully explain specificity. Solution biochemical and computational studies have shown that a comprehensive view of specificity determination must also include factors such as shape recognition (2), mutual accommodation of the macromolecules through DNA distortion or conformational selection (1, 3–6), and/or DNA-induced protein folding (5, 7). Thermodynamic studies have revealed that specific protein-DNA association is often driven primarily by the favorable entropy increase provided by desolvation of the apposed complementary surfaces (8–10).

For those DNA-binding proteins that are also DNA-modifying enzymes (nucleases, methylases, recombinases, repair enzymes, etc.) a key question is the relationship between DNA-binding specificity and catalytic specificity. One exemplary system for addressing specificity determination is the EcoRI endonuclease (3, 11), a 62 kDa homodimer that recognizes the DNA site 5′-GAATTC-3′ and binds as much as 90,000-fold better (3, 12) than at sites that differ by one base pair from the specific site. Full binding specificity is achieved in the absence of divalent metal (3, 12, 13).

Like most phosphotransferases and all metallonucleases, EcoRI requires a divalent metal as an obligatory participant in catalysis. The fully symmetrical complex of EcoRI homodimer with its palindromic cognate DNA contains two catalytic centers, each formed from elements of both subunits. In each center, one Mg²⁺ coordinates with a pair of carboxylate side chains and the scissile phosphate GpAATTC (14, 15). Catalytic rates for EcoRI vary according to the series Mg²⁺ ≈ Mn²⁺ > Co²⁺≫Zn²⁺≫Cd²⁺ > Ni²⁺ (16, 17). Ca²⁺ cannot support catalysis, but can act as an inhibitor by competing with the physiological cofactor Mg²⁺ (15).

Based on studies with stereospecific phosphate analogues and molecular dynamics simulations, Kurpiewski et al. (15) proposed a model of EcoRI catalysis in which the phosphate at GApATTC (one step downstream from the scissile phosphate GpAATTC) precisely orients the attacking nucleophilic water through an intermediate “relay” water molecule. This phosphate is uniquely positioned by the “kinked” DNA distortion in the fully correct recognition complex (15). The postulated role of sequence-dependent DNA distortion may account for the fact that catalytic rate depends on recognition of the correct DNA site: When Mg²⁺ is added to prebound protein-DNA complexes, the catalytic rates for the chemical steps of double-strand DNA cleavage are 10⁶-fold higher for the correct GAATTC DNA sites than for sites that differ from GAATTC by as little as one base pair (3, 15). That is, only the fully correct recognition complex supports the maximum catalytic rate. The multiplicative binding and catalytic specificities enable EcoRI to distinguish a single target site from the competing excess of near-correct DNA sites in vivo (3, 9, 12).

Here we combine electron spin resonance (ESR) spectroscopy and biochemical studies to study the locations and functional consequences of paramagnetic Cu²⁺ ions bound at hitherto unknown sites in the EcoRI endonuclease-DNA complex. Cu²⁺ ions, which do not support DNA cleavage (16), have the potential to provide fixed points of reference for ESR studies of the structure and dynamics of these complexes and other phosphotransferases. We used continuous wave (CW) ESR and electron spin echo envelope modulation (ESEEM) to probe the interaction of the Cu²⁺ electron spin with surrounding nuclei. These experiments established that Cu²⁺ ions are coordinated to EcoRI histidines. Isothermal titration calorimetry (ITC) determined a stoichiometry of two Cu²⁺ bound per EcoRI dimer-DNA complex. We identified His114 (one in each monomer) as the site of Cu²⁺ binding by double electron-electron resonance (DEER)-ESR measurements of both Cu²⁺-Cu²⁺ and Cu²⁺-nitroxide distances. We confirmed this identification by showing that the His114Tyr mutation abolished the effect of Cu²⁺ on both DNA binding and DNA cleavage.

Unexpectedly, we found that Cu²⁺ inhibits the chemical step of DNA cleavage, despite binding at sites 13 Å away from those at which Mg²⁺ binds. Molecular dynamics simulations suggest that upon Cu²⁺ binding at His114, protein side chains and water molecules that are important in catalysis are displaced and the crucial position of the 3′-neighboring DNA phosphate is altered. This model differs from any previously proposed for metal inhibition of a metallonuclease. We also show that there is positive cooperativity in Mg²⁺ association with the EcoRI-DNA complex to promote catalysis, and we propose an electrostatic basis for this interaction between catalytic sites. Positive cooperativity between catalytic sites may help ensure that EcoRI fulfills its biological function of inactivating foreign DNA by irreparable double-strand DNA cuts, in preference to single-strand nicks that are readily repaired by DNA ligase.

Results and Discussion

ESEEM Shows Cu²⁺ Is Coordinated to a Histidine.

The signal from Cu²⁺ binding to the EcoRI-DNA complex is readily detected in the CW-ESR spectrum (Fig. 1A). There are two distinct spectral components, and the best-fit simulated spectrum (black dotted lines, Fig. 1A) is obtained when the ratio of these two components is approximately 1∶1. The magnetic g_‖ and A_‖ tensor values are within the range found for type-II Cu²⁺ complexes (18), indicating that both Cu²⁺ components have four equatorial ligands. The values of the magnetic g_‖ and A_‖ tensors of the first component are consistent with either one, two, or three nitrogen ligands, with oxygens providing the remaining ligands (18). The second component has a smaller A_‖ value, which can be attributed to distortion in the planarity of the four equatorial ligands (19). Evidence from distance measurements combined with experimental determination of stoichiometry (below) shows that a single Cu²⁺ ion binds at the same location in each subunit; thus, the two components do not arise from heterogeneity in the locations of bound Cu²⁺ ions. We further address the origin of two components below in the context of ITC, DEER, and molecular dynamics (MD) results.

Fig. 1.

ESR data on the Cu²⁺ bound EcoRI-DNA complex. (A) The CW-ESR spectrum at 80 K (gray solid line) shows two components. The second component is clear in the magnified section of the spectrum, which is shown in the Inset. The simulated spectrum is shown by the dotted line. The best-fit simulation was obtained with the following parameters: component 1: g_‖ = 2.289, A_‖ = 163 G, g_⊥ = 2.06, A_⊥ = 20 G and component 2: g_‖ = 2.228, A_‖ = 143 G, g_⊥ = 2.06, A_⊥ = 20 G. The relative ratio of the two components is approximately 1∶1. (B) The three-pulse ESEEM spectrum at 20 K obtained at a magnetic field of 3369 G. The peaks between 0–2 MHz, indicated by the trident, are assigned to the imidazole ¹⁴N from a histidine residue. The broad peaks at approximately 4.0 MHz are assigned to the double quantum transition. The peak at approximately 14.3 MHz is assigned to the proton ESEEM peak. (Inset) The Cu²⁺ coordination derived from the ESEEM results.

To further characterize the Cu²⁺ coordination, we obtained a three-pulse ESEEM spectrum at a magnetic field that corresponds to the g_⊥ region of the Cu²⁺ ESR spectrum, so that both Cu²⁺ components contribute to the ESEEM signal. The ESEEM spectrum (Fig. 1B) has two sharp peaks (approximately 0.60 MHz and approximately 1.60 MHz) and one broad shoulder peak (approximately 1.09 MHz). The sum of the two lower frequencies is close to the highest one, indicating that they are mainly due to the nuclear quadrupole interactions of a weakly coupled ¹⁴N (20–22). Both the quadrupole parameters derived from these frequencies (e²Qq/h approximately 1.73 MHz and η approximately 0.69) (22) and the three peaks in the ESEEM spectrum are typical of remote, noncoordinated ¹⁴N from a histidine imidazole ring bound to Cu²⁺ (20, 21, 23–27). Broad peaks at approximately 4.0 MHz are also resolved (Fig. 1B). These peaks may arise from the double quantum transition of ¹⁴N (21). Hyperfine sublevel correlation experiments (Fig. S1) show characteristic cross-peaks that support the assignment that all the peaks at approximately 4.0 MHz in the ESEEM spectrum are from the double quantum transition. The peak at 14.3 MHz in the ESEEM spectrum is from weakly coupled protons around the Cu²⁺ electron spin. These protons may come from either the solvent or the macromolecules.

Strategy for Identifying the Cu²⁺ Binding Histidine Residue.

There are five His residues in each subunit of EcoRI, at positions 31, 114, 147, 162, and 225. We employed a four-stage strategy to identify the specific His residues that bind Cu²⁺ in the EcoRI-DNA complex. First, we used DEER-ESR to measure the distance between Cu²⁺ ions, and compared this to inter-His distances in the cocrystal structure. The single distance obtained narrowed the field of candidate binding sites. Second, we made a Ser180Cys mutant and nitroxide spin-labeled the Cys residues, and then used DEER-ESR to measure Cu²⁺-nitroxide distances. The data yielded two Cu²⁺-nitroxide distances (intersubunit and intrasubunit), from which triangulation unambiguously identified His114 as the Cu²⁺-binding residue. Third, the binding site was confirmed by showing that a His114Tyr mutation completely abolishes the ability of Cu²⁺ to stimulate protein-DNA binding and to inhibit DNA cleavage. Fourth, we showed by ITC a stoichiometry of two Cu²⁺ ions bound per EcoRI dimer.

Cu²⁺-Cu²⁺ Distance Excludes Some Candidate Binding Sites.

To narrow the set of candidate Cu²⁺ binding sites, we used the DEER experiment to measure the distance between the two Cu²⁺ ions (one in each EcoRI subunit). DEER data were collected at five magnetic fields and a fixed resonance offset (Fig. 2 A), and also with multiple resonance offsets (Fig. S2B). The field and/or resonance offset variation of the DEER signal provides information about Cu²⁺-Cu²⁺ distance as well as the relative orientation of the two Cu²⁺ centers (28–30). The data showed only very weak orientational effects, as described in the legend in Fig. 2.

Fig. 2.

Cu²⁺-Cu²⁺ distance measurements on the EcoRI-DNA complex. (A) Baseline corrected DEER data at five different magnetic fields with a fixed resonance offset of 100 MHz, obtained at 20 K. The positions of the observed pulse are highlighted on the field-swept echo detected spectrum shown in the Inset—the exact magnetic fields are provided on each experimental trace. The fast modulations in data collected at 3240 G and 3190 G are due to proton ESEEM effects. No significant magnetic field or resonance offset dependence of the baseline-corrected signal was observed at the g_⊥ region. The modulation period of the data collected at the g_‖ region (at 3090 G) deviated from the other datasets by approximately 60 ns (see vertical dotted line in Fig. 2A), showing very weak orientational selectivity. Simulated data are shown by dashed lines. (B) The measured Cu²⁺-Cu²⁺ distance distribution function. (C) The crystal structure showing the intersubunit His-Nε to His-Nε distances that are closest to the experimental distance.

To obtain the Cu²⁺-Cu²⁺ distance from the DEER signal, a simple molecular model (Fig. S2A) and the fitting procedure developed in our recent work (30) were applied to this complex. The simulated signals using the optimized parameter set are shown by the dashed lines in Fig. 2A. The distance distribution obtained from the optimized parameter set (Fig. 2B) showed a single most probable Cu²⁺-Cu²⁺ distance of 35 ± 1 Å (mean ± SD). Details of analysis and specificity of parameters are provided in Figs. S2 and S3.

We then compared the DEER-derived Cu²⁺-Cu²⁺ distance to all possible inter-His distances (Table S1) from the crystal structure of the metal-free EcoRI-DNA complex (14, 31). This comparison suggests that either His114-Nε to His114-Nε (33 Å) or His162-Nε to His162-Nε (32 Å) distances (Fig. 2C) might be consistent with the DEER data. The His225-Nε to His225-Nε distance (28 Å) is less consistent with the DEER measurements, although the single distance, by itself, is insufficient to conclusively exclude His225. On the other hand, symmetrical binding sites at His31 (Nε-Nε) (69 Å) or His147 (Nε-Nε) (10 Å) are plainly excluded. Binding to two different histidines in each subunit of the dimer (e.g., His31-Nε to His147-Nε at 38 Å or His114-Nε to His225-Nε at 36 Å) is intrinsically unlikely because of the symmetry of the homodimer (14), and is in any event excluded by triangulation with Cu²⁺-nitroxide distances (below) and ITC results.

Triangulation with Cu²⁺-Nitroxide Distances Shows That Cu²⁺ Coordinates at His114.

The distance(s) between Cu²⁺ and a known point on the protein provides additional constraints for triangulation. We mutated Ser180 to Cys and labeled the Ser180Cys mutant protein with the methanethiolsulfonate nitroxide spin-label (MTSSL) (32). The Ser180Cys mutation causes minimal perturbation to the protein structure and the spin-labeled EcoRI retains normal DNA-binding affinity (32). To selectively measure the Cu²⁺-Ser180Cys distances, DEER signals for the Ser180Cys-DNA-Cu²⁺ complex were collected with the pump pulse applied to the nitroxide ESR spectrum and the observer pulses applied to the Cu²⁺ ESR spectrum (Fig. 3A, Inset). Internal orientation effects were investigated by collecting DEER data at different frequency offsets (Fig. 3A). Two clear modulation periods were observed, indicating a bimodal distribution of Cu²⁺-nitroxide distances. Analysis of the data yielded a bimodal distance distribution function (Fig. 3B), with the most probable distances at 22 ± 2 Å and 42 ± 3 Å (mean ± SD). Details of analysis and specificity of parameters are provided in Fig. S4.

Fig. 3.

Cu²⁺-nitroxide distance measurements on the MTSSL-Ser180Cys-EcoRI-DNA-(Cu²⁺)₂ complex. (A) Baseline corrected DEER data at four different resonance offsets, 266 MHz, 100 MHz, 448 MHz, and 548 MHz (from Top to Bottom), obtained at 20 K. The positions of the pump and observer pulses are indicated on the field-swept echo detected spectrum shown on the inset. Simulated data are shown by dashed lines. (B) The measured Cu²⁺-nitroxide distance distribution function. (C) The intersubunit and intrasubunit distances between candidate His-Nε and residue 180-Cα. (D) Summary of triangulation results, showing Cu²⁺-Cu²⁺ distance (red) and Cu²⁺-nitroxide distances (green).

The crystal structure of the metal-free EcoRI-DNA complex shows that His114-Nε is 21 Å from the Cα of residue 180 on the same monomer, and about 42 Å from 180-Cα on the other monomer (Fig. 3C, solid green line). These values are very close to those from the DEER-ESR analysis. The corresponding distances for His162 are 41 Å and 47 Å, and His225 are 45 Å and 51 Å (Fig. 3C, dashed black and dotted blue lines); thus, the Cu²⁺-nitroxide constraints exclude these two candidates. The triangulation of Cu²⁺-Cu²⁺ and Cu²⁺-nitroxide distance constraints also excludes all possible modes of metal binding to two different His residues in a subunit. The experimental DEER results are therefore uniquely consistent with the hypothesis that His114 binds Cu²⁺ in each subunit (Fig. 3D; Tables S1 and S2).

Mutation Confirms Cu²⁺ Binding at His114.

The interaction of Cu²⁺ ions with the EcoRI-DNA complex was measured by the dependence of protein-DNA affinity on Cu²⁺ concentration. Interestingly, Cu²⁺ ions increase protein-DNA binding affinity about eightfold (Fig. 4A). Divalent metals generally promote protein-DNA binding by decreasing the charge repulsion between the protein and nearby DNA phosphates (15, 33, 34). For example, we previously showed that the competitive inhibitor Ca²⁺ binds to EcoRI at the negative charge cluster in the active site and thereby increases the equilibrium association constant (K_A) as much as 380-fold (15). Cu²⁺ evidently has a weaker effect than Ca²⁺ because the Cu²⁺ (coordinated at His114) lies at greater distance (10 Å) from the nearest DNA phosphate GApATTC.

Fig. 4.

Interaction of Cu²⁺ with the EcoRI-DNA complex. (A) Cu²⁺ enhancement of EcoRI binding to cognate duplex TCGCGAATTCGCG (recognition site underlined). Ratio of K_A (plus Cu²⁺) to K_A (no Cu²⁺) is plotted for wild-type EcoRI (empty circle) and His114Tyr (solid circle) enzymes; data show means ± SD for at least three determinations at each [CuCl₂] or at least six determinations without Cu²⁺. For wild-type EcoRI, K_A,0(no Cu²⁺) = 1.2( ± 0.1) × 10⁹ M^-1; K_A,Cu²⁺(max ) = 9.3( ± 0.3) × 10⁹ M^-1. For the promiscuous His114Tyr that shows better binding affinity than wild-type EcoRI (12), K_A,0 = 1.7( ± 0.1) × 10¹⁰ M^-1 and K_A,Cu²⁺(max ) = 7.3( ± 2.1) × 10¹⁰ M^-1. Fit to the Hill equation K_A,Cu²⁺ = K_{A,Cu²⁺(max )}[Cu²⁺]ⁿ/(K_{D 0.5,Cu²⁺} + [Cu²⁺]ⁿ) gives Hill coefficient n = 1.0 ± 0.15 and K_{D 0.5,Cu²⁺} = 5.1 ± 0.8 μM. Inset shows the hyperbolic binding isotherm of wild-type EcoRI for [Cu²⁺] from 0 to 100 μM. (B) Representative ITC titration of CuCl₂ (1 mM) into sample cell (15 °C) containing EcoRI (9 μM) and DNA (20 μM). Top Box shows heat signals obtained for 125 injections (2.5 μL each). Lower Box shows integrated heat signal normalized to moles of Cu²⁺ injected (after subtraction of heats of dilution of Cu²⁺) per mole of EcoRI-DNA complex. The stoichiometry value n = 2.0 ± 0.3 and K_D value of 14 ± 6 μM (means ± SD of at least three determinations at 10 °C, 15 °C, 21 °C) did not vary with temperature or buffer identity (imidazole or TRIS; see Materials and Methods for buffer compositions).

When the putative Cu²⁺ binding residue is removed in His114Tyr mutant EcoRI, the stimulation by Cu²⁺ is almost completely abolished (Fig. 4A). The apparent affinity for Cu²⁺ is reduced by 1,600-fold by the mutation; compare apparent K_D,Cu²⁺ = 5.1 ± 0.8 μM for wild-type EcoRI to K_D,Cu²⁺ approximately 8 mM for the His114Tyr mutant. These observations confirm that Cu²⁺ ion binds at His114, a site distinct from that where Mg²⁺ binds.

Direct Determination of Cu²⁺ Binding Stoichiometry by ITC.

The ESR data above give a single Cu²⁺-Cu²⁺ distance in the complex, suggesting one Cu²⁺ is bound to each EcoRI subunit. Detection of only two Cu²⁺-nitroxide distances (one intersubunit and one intrasubunit) is also indicative of only one Cu²⁺ bound per subunit. To confirm this inference, we determined stoichiometry directly by calorimetric measurement of the heat of binding. Cu²⁺ was titrated into a solution of the wild-type EcoRI-DNA complex (Fig. 4B). The parameters obtained were K_D,Cu²⁺ = 14 ± 6 μM, and a stoichiometry of 2.0 ± 0.3 Cu²⁺ bound per EcoRI-DNA complex. The ITC results thus confirm that each subunit of the EcoRI dimer coordinates one Cu²⁺ ion.

Cooperative Mg²⁺ Interactions in the Two Catalytic Sites.

To provide a sound basis for understanding Cu²⁺ effects on catalysis (see below), we carefully reinvestigated the interaction of Mg²⁺ with the EcoRI-DNA complex. In contrast to some metallonucleases that use two Mg²⁺ ions per catalytic site, EcoRI is generally considered (14, 17, 35) to use only a single Mg²⁺ in each catalytic site. Each Mg²⁺ is coordinated to Glu111, Asp91, and GpAATTC (see last section of Results and Discussion). On the basis of cocrystal structures (14) and MD simulations (15), one phosphoryl oxygen (O1P) at the scissile GpAA is coordinated by Mg²⁺ while the other phosphoryl oxygen (O2P) hydrogen bonds to positively charged side chains of Lys113 and Arg145; these side chains thus mimic both the position and the functional role of the second metal ion proposed for many two-metal-ion phosphotransferase reactions, namely to stabilize the phosphorane transition state and/or the oxyanion leaving group (36–40).

To determine effects on catalysis independent of binding effects (3), we preformed EcoRI-DNA complexes in the absence of Mg²⁺ under conditions of molar excess of enzyme to DNA, and then initiated reaction by the addition of Mg²⁺ at various concentrations. The resulting single-turnover first-order rate constants, measured separately for each DNA strand, thus represent a composite of Mg²⁺ binding to the EcoRI-DNA complex (12) and the chemical step of catalysis. The rate constants for wild-type EcoRI cleavage of the two DNA strands show sigmoidal dependence on [Mg²⁺] (Fig. 5 A); the best fits of the kinetic data to the Hill equation k_cleave = k_max[Mg²⁺]ⁿ/(K_Mg²⁺ + [Mg²⁺]ⁿ) yield values of n = 1.9 ± 0.1 and n = 1.8 ± 0.1 for cleavage of each of the two DNA strands. The slopes of log-log Hill plots (Inset, Fig. 5A) for the kinetic data give similar values. The observed value of n = 1.8–1.9, given that EcoRI uses one metal ion per catalytic center (11, 14, 17, 41), implies interaction between the single-Mg²⁺ binding sites of the protein homodimer.

Fig. 5.

Dependence of single-turnover cleavage rate constants on Mg²⁺ concentration. Y-axis (k_cleave) represents first-order rate constants k₁ (empty circle) or k₂ (empty triangle) for cleavage of the distinguishable top and bottom DNA strands (Materials and Methods). Data were fit to the Hill equation (see text). (A) The apparent affinities for Mg²⁺ are K_0.5,Mg²⁺ = 2.4 ± 0.1 mM for wild-type, measured for both k₁ and k₂ and (B) K_0.5,Mg²⁺ = 2.2 ± 0.2 mM for His114Tyr enzyme. For wild-type EcoRI, Hill coefficient n values were 1.9 ± 0.1 (k₁) and 1.8 ± 0.1 (k₂) and for His114Tyr, n values were 2.0 ± 0.3 (k₁) and 1.9 ± 0.3 (k₂). Each point is the mean ± SD of at least three determinations.

Positive cooperativity for Mg²⁺ interaction has also been observed for restriction endonuclease PvuII, which binds two metal ions in each of its two catalytic sites (42, 43); Hill coefficients of n = 3.6 to 4 were reported (44, 45). Although the precise role of the second metal ion is in general controversial for many metallonucleases (17, 35, 39, 40), for PvuII the first Mg²⁺ bound in a catalytic site is both necessary and sufficient for catalysis (45). The second Mg²⁺ acts as a powerful accelerant for DNA cleavage. This accelerant function, in itself, would produce intrasite cooperativity when cleavage rate is used as a metric, but additional intersite interaction would have to be postulated to account for the large Hill coefficient. In contrast, our data showing Hill coefficient n ≈ 2, when EcoRI has only one Mg²⁺ in each catalytic site, necessarily imply intersite cooperativity.

Mg²⁺ Cooperativity Reflects Long-Range Electrostatics.

It was not obvious from any structural considerations why two single-Mg²⁺ binding sites approximately 20 Å apart should show positive cooperativity. To explore this phenomenon, we first carried out multiple parallel MD simulations of the EcoRI-DNA complex in the presence of two Mg²⁺ ions, or only one Mg²⁺ ion. As in previous MD studies (15), introduction of Mg²⁺ into both catalytic sites caused side-chain movements. Asp91 and Glu111 rotate to coordinate Mg²⁺ and moves (from an initial position approximately 5 Å away, also its equilibrium position in Mg²⁺-free simulations, Fig. S5) to form a stable interaction with the scissile phosphate GpAATTC at approximately 2.7 Å. We now did the in silico experiment of placing Mg²⁺ in only one of the two catalytic sites, and observed (trajectories Fig. S5) that this caused not only the expected movement of the Lys113 side chain in the Mg²⁺-containing site but also, unexpectedly, a similar movement of Lys113 in the distal, Mg²⁺-free catalytic site to form a stable interaction with the scissile phosphate GpAATTC. This result suggested that binding of the first Mg²⁺ might produce effects at considerable distance.

In pioneering papers (46, 47), Honig and coworkers showed that macromolecules can generate functionally important patterns of surface electrostatic potential that depend not only on the locations of charged groups, but also on the shape of the molecular surface. Regions of unusual negative potential in particular can define functionally significant metal binding sites (48). We therefore calculated electrostatic surface potentials with the DelPhi program (49, 50), that provides finite-difference solutions to the nonlinear Poisson-Boltzmann equation (51). These calculations show not only the expected increase in positive potential upon Mg²⁺ addition to one site, but also that the potential over a large region surrounding the Mg²⁺-free active site approximately 20 Å away becomes pronouncedly more negative than it was in the metal-free complex (Fig. S6). (At the specific “dummy atom” position corresponding to that of the absent Mg²⁺, the change was about -20 kT/e, Table S3.) This in silico experiment was varied by placing one Mg²⁺ in the opposite catalytic site, and again the Mg²⁺-free catalytic site developed a more negative potential, to nearly the same extent (Table S3). Thus, binding of Mg²⁺ in one catalytic site causes the electrostatic potential to become more negative in the distal metal-free site. This enhancement of negative electrostatic potential should promote binding of the second Mg²⁺, thus providing a sufficient explanation for the observed positive cooperativity. In a functional sense, positive cooperativity in Mg²⁺ binding would tend to promote double-strand DNA cleavage over single-strand nicking, as required for a restriction endonuclease to introduce inactivating double-strand breaks into foreign DNA.

Cu²⁺ Inhibits Mg²⁺-Dependent Catalysis.

To examine the catalytic consequences of Cu²⁺ binding, we performed single-turnover assays to measure rate constants for the chemical step. Significantly we found that Cu²⁺ is a powerful inhibitor of DNA cleavage by wild-type EcoRI (Fig. 6). This inhibition is unexpected because Cu²⁺ does not bind directly at the catalytic site. Because this single-turnover assay uses preformed EcoRI-DNA complexes with molar excess of enzyme over DNA, Cu²⁺ must affect the chemical step rather than protein-DNA association or product release. Note that the apparent Cu²⁺ affinity as an inhibitor of catalysis (K_0.5,Cu²⁺ ≈ 70 μM) is weaker than the affinity in the absence of Mg²⁺ from calorimetry (14 μM, Fig. 4B) or the apparent affinity from equilibrium binding stimulation (5 μM, Fig. 4A), also in the absence of Mg²⁺. This difference likely reflects Mg²⁺-Cu²⁺ repulsion in the cleavage assays, at an inter-ion distance of approximately 13 Å (see Fig. 7A below). This difference in Cu²⁺ affinities is also responsible for the sigmoid shape of the Cu²⁺ inhibition curve.

Fig. 6.

Inhibition of EcoRI cleavage by Cu²⁺. See procedure in Materials and Methods. Ratio of first-order cleavage rate constant k_cleave(+Cu²⁺)/k_cleave(no Cu²⁺) as a function of [CuCl₂]. Wild-type EcoRI cleavage reactions were performed at Mg²⁺ concentrations 3 mM (upside down triangle), 4 mM (empty triangle), 6 mM (empty square), and 8 mM (empty circle); His114Tyr reaction was at 8 mM (solid circle). Plots for k₁ (cleavage in top strand) and k₂ (cleavage in bottom strand) were precisely superimposable. For clarity, only k₁ data are shown. Each point represents the mean of at least three determinations. Error bars are too small to be seen at this scale.

Fig. 7.

Simulation of Cu²⁺-induced structural changes in the catalytic site of the EcoRI-DNA complex. Superimposed images of EcoRI•DNA•(Mg²⁺)₂ complex (protein side chains green; DNA cyan) and the EcoRI•DNA•(Mg²⁺)₂•(Cu²⁺)₁ complex (protein side chains orange; DNA magenta). Structures from MD snapshots (see Materials and Methods) were matched at the DNA nucleotides CGA using Chimera software (59). Hydrogen bonding distances (yellow dashed lines and numbers) are given in Ångstroms. (A) In the complex without Cu²⁺, a water molecule (W_A) coordinated to Mg²⁺ carries out nucleophilic attack (white arrow) at the scissile phosphate GpAATTC. This water is hydrogen-bonded to the stable “relay water” (W_C), which is precisely positioned by H-bonding to the adjacent phosphate GApATTC, while His114-Nδ H-bonds to the phosphate GApATTC. In the Mg²⁺ + Cu²⁺ complex, His114 and GApATTC assume completely different positions and there is no stable equivalent of W_C (see text). (B) Rotated view of the region, showing how Arg145-guanidino in the Mg²⁺ + Cu²⁺ complex makes H-bonds (double-headed arrows) to GpAATTC and GApAATTC (thus bridging both phosphates), while retaining an H-bond to recognize Ade-N7. Within < 2 ps after the start of the simulation for the EcoRI•DNA•(Mg²⁺)₂ complex, stable (± 0.1 Å) coordination distances to Mg²⁺ were achieved as follows: GpA-O1P, 1.85 Å; Asp91-OD1/OD2, 1.9 Å, Glu111-OE2, 1.9 Å, W_A, 1.95 Å. An additional coordination fluctuates between Ala112-carbonyl (3.0 ± 0.5 Å) and a transiently visiting water.

Strikingly, for the His114Tyr mutant EcoRI, Cu²⁺ (even at 200 μM) does not inhibit the Mg²⁺-catalyzed cleavage reaction, whereas 100 μM Cu²⁺ completely inhibits catalysis by wild-type EcoRI (Fig. 6). As a control, we showed that the His114Tyr mutation does not affect affinity for Mg²⁺ or the cooperativity of Mg²⁺ action (Fig. 5B). The insensitivity of His114Tyr protein to Cu²⁺ inhibition thus confirms our inference from ESR and the binding assays (see above) that His114 is the coordinating site for Cu²⁺.

It is interesting that the effective Cu²⁺ concentrations for inhibition of wild-type enzyme are approximately equivalent to normal Cu²⁺ concentrations in vivo (52), suggesting that coordination to histidine sites in proteins could have a biologically significant role in Cu²⁺ toxicity.

How Does Cu²⁺ Inhibit EcoRI Catalysis?

In the wild-type EcoRI-DNA complex, His114 donates a hydrogen bond to the phosphate at GApATTC. In the proposed catalytic mechanism (15) in the absence of Cu²⁺, this phosphate serves to position a water molecule (W_C in Fig. 7A), that in turn hydrogen bonds to the Mg²⁺-bound water that carries out nucleophilic attack at GpAATTC. In addition, studies with chiral phosphate analogs (15) implied that DNA distortion in the complex is required for catalysis and that the precise positioning of the phosphoryl oxygen at GApA that participates in the “water relay” is important for the rate of cleavage. This preorganized arrangement is thus an example of what Bruice and Benkovic (53) termed a “near-attack conformer.”

To investigate how Cu²⁺ inhibits catalysis, we carried out MD simulations of the EcoRI-DNA-(Mg²⁺)₂ complex, with or without Cu²⁺ (see below). Coordination of the Cu²⁺ ions at His114-Nε yields a stabilized mean intersubunit Cu²⁺-Cu²⁺ distance of 38 Å, close to the 35 Å from ESR-DEER; both ESR and MD distributions have distribution widths at half-height approximately 2–3 Å.

Detailed MD simulations with explicit solvent show that when one Cu²⁺ is inserted into the EcoRI•DNA•(Mg²⁺)₂ complex at His114-Nε (see above), His114 assumes a position in which Cu²⁺ is coordinated by His114-Nε, Glu170 and Glu245; this requires rotations of the Glu170 and Glu245 side chains (Fig. 7A). The simulations suggest that transient rotation of the Glu170 side chain allows both carboxylate oxygens to coordinate alternately to the Cu²⁺ ion. This rotation can create distortions in the planarity of the equatorial ligands and account for the second spectral component (with small A_‖ value) that we observed in CW-ESR.

After these side chain adjustments, His114 is approximately 4 Å from its normal position. (No comparable repositioning occurs in the Cu²⁺-free catalytic site, suggesting that the Cu²⁺ binding sites in each subunit have no mutual influence at a range of 36 Å. Similar movement occurs in both catalytic sites when Cu²⁺ ions are inserted in both.) Lacking interaction with His114-Nδ, the GApA phosphate rotates to where it cannot support the water network (Fig. 7A). The GApA is stabilized in this abnormal position by a newly formed interaction with Arg145 (Fig. 7B). Arg145 remains in position to donate hydrogen bonds to the scissile GpAA (15), and simultaneously to N7 of the inner adenine (Fig. 7B). Arg145 was proposed (15) to polarize the scissile phosphate and/or to stabilize the phosphorane intermediate in the transition state; this function may be compromised by the bridging of one Arg145 guanidino nitrogen between the GpAA and GApA phosphates (double arrows, Fig. 7B) upon addition of Cu²⁺. Crucially, the water relay (water C, Fig. 7A) from the GApA to the nucleophilic water no longer exists in stable form. In the absence of Cu²⁺, this position has 98–100% occupancy (in various parallel simulations) with average residence times for the waters in the range 300 to 500 ps, but in the Cu²⁺-bound site it is only 20–40% occupied (in various parallel simulations) by waters with short residence times (< 100 ps). The overall picture is that the “near-attack conformer” essential for catalysis no longer exists in the Cu²⁺ complex.

Our findings therefore strengthen the view that coupling between recognition and catalytic specificity for EcoRI derives from (i) A role of DNA distortion in both recognition and catalysis; (ii) the use of functional groups (e.g., Arg145, Fig. 7B) for dual roles in sequence recognition and the catalytic mechanism; (iii) and communication between the two active sites. It will be intriguing to learn if additional principles have been evolved by other site-specific DNA-modifying enzymes.

Materials and Methods

Enzyme Expression and Purification.

The EcoRI endonuclease gene was expressed as a fusion with maltose-binding protein and cleaved after purification to produce the complete EcoRI protein with no extra amino acids. Details are in SI Text.

Oligodeoxynucleotide Substrates.

Oligodeoxynucleotides (Integrated DNA Technologies) were purified as described previously (12). The self-complementary TCGCGAATTCGCG (12 bp with dangling T) was used for ESR and equilibrium binding experiments (see below). For cleavage rate assays, duplex deoxyoligonucleotide, (23 base-pairs plus dangling end base), was annealed from purified single-stranded 24 nt top (5′-GATGGGTGCAGAATTCTGCAGGTA-3′) and bottom (5′-CTACCTGCAGAATTCTGCACCCAT-3′) sequences as described by Sapienza, et al. (12). For both equilibrium binding and cleavage experiments, the DNA was 5′-end labeled with ³²P-ATP using T4 polynucleotide kinase as described (54). The substrate was designed such that cleavage in the top and bottom strands produces 5′-³²P-labeled 11 nt and 10 nt products, respectively.

Preparation of Cu²⁺-Wt EcoRI-DNA and Cu²⁺-S180C-MTSSL-DNA ESR Samples.

A solution of wild-type EcoRI (5 μM) in the presence of fivefold molar excess of TCGCGAATTCGCG was exchanged into 30 mM N-ethylmorpholine (NEM) buffer containing 0.3 M NH₄Cl, 10% dioxane, 30% deuterated glycerol (d8), 65% D₂O (pH 8.0) and concentrated. The final concentrations of EcoRI and 13mer DNA were 380 μM and 1.5 mM, respectively.

Details of nitroxide spin-labeling are in SI Text. The final concentrations of Ser180Cys protein and 13mer DNA were 180 μM and 1.8 mM respectively. For both Cu²⁺-EcoRI-DNA and Cu²⁺-MTSSL-S180C-DNA samples, isotopically enriched (Cambridge Isotope Labs, Inc) was added at a 4∶1 molar ratio (Cu²⁺: protein dimer). The samples were stored at -80 °C and flash frozen to 20 K before each DEER ESR experiment.

ESR Experiments.

All of the pulsed ESR experiments were performed on a Bruker Elexsys 580 spectrometer at 20 K. CW-ESR and ESEEM experiments were performed with a MS3 resonator. DEER experiments were performed with a MD5 resonator. Details of pulse sequences, data acquisition, and analysis are in SI Text.

Equilibrium Binding Studies.

Equilibrium association constants (K_A) of Ser180Cys protein and MTSSL-Ser180Cys protein to DNA were determined as described (32). Experiments assessing Cu²⁺ enhancement of binding used the same double-strand deoxyoligonucleotide TCGCGAATTCGCG as that in the ESR experiments. Equilibrium association constants (K_A) for wild-type EcoRI and His114Tyr (± Cu²⁺) were measured by the nitrocellulose filter binding assay as described (34, 55). For assays in the absence of Cu²⁺, EDTA (1 mM) was included in the binding buffer (30 mM NEM, 10% glycerol, 10% dioxane, 100 μM dithiothreitol, 100 μg/mL bovine serum albumin) plus 0.3 M KCl; pH 8.0, 22 °C.

Cleavage Kinetics.

Single-turnover cleavage rates (3, 12) using ³²P-labeled 23 base-pair DNA (see above) were determined as a function of Mg²⁺ concentration by preforming EcoRI-DNA complexes (fivefold molar excess of EcoRI over DNA) and initiating the reaction by addition of Mg²⁺. Inhibition by Cu²⁺ was determined by two different protocols (SI Text) that gave identical results.

Isothermal Titration Calorimetry.

ITC experiments were performed by repeated injections of CuCl₂ into a solution of EcoRI-DNA complex until complete saturation, where there was no further change in the heat of reaction. Observed heats of reaction were corrected for the heat of dilution. Details of procedure and data analysis are in SI Text.

Molecular Dynamics Simulations.

MD simulations using the combined all-atom CHARMM22/CMAP and CHARMM27 force fields (56–58) were performed on EcoRI-TCGCGAATTCGCG complexes with (i) no added metal ion, (ii) one active site (either A or B) with Mg²⁺, (iii) both active sites A and B with Mg²⁺, (iv) one subunit (either A or B) with Cu²⁺, (v) both subunits with Cu²⁺, (vi) both active sites A and B filled with Mg²⁺, one Cu²⁺ added to one subunit (either A or B), (vii) both active sites A and B filled with Mg²⁺, and a Cu²⁺ ion added to both subunits. The starting point for all simulations was a highly refined, high-resolution (1.85 Å) crystal structure model (14) based on PDB ID code 1CKQ. For each case (with or without Mg²⁺ and/or Cu²⁺), multiple parallel production runs (ranging from 4 ns to 10 ns) were performed. Details are in SI Text. Molecular graphics images were produced using the UCSF Chimera package (59) from the Resource for Biocomputing, Visualization, and Informatics at the University of California, San Francisco.

Electrostatic Potential Calculations.

Electrostatic potentials were calculated and mapped to molecular surfaces using an updated version (v.5) of the DelPhi program which uses finite-difference methods to solve the nonlinear Poisson-Boltzmann equation (FDPB method) (49, 50). Input coordinate files for the DelPhi calculations were generated from snapshots from multiple parallel molecular dynamics simulations (see above) of the EcoRI-DNA complex in the absence of Mg²⁺ or with one Mg²⁺ added to either active site A or B. Details are given in SI Text. Electrostatic potential maps of the molecular surfaces of the EcoRI-DNA complexes were visualized using PyMOL (60).