Role of active site loop dynamics in mediating ligand release from E. coli dihydrofolate reductase

Abstract

Conformational fluctuations from ground-state to sparsely populated but functionally important excited states play a key role in enzyme catalysis. For Escherichia coli dihydrofolate reductase (DHFR), release of the product tetrahydrofolate (THF) and oxidized cofactor NADP⁺ occurs through exchange between closed and occluded conformations of the Met20 loop. A “dynamic knockout” mutant of E. coli DHFR, where the E. coli sequence in the Met20 loop is replaced by the human sequence (N23PP/S148A), models human DHFR and is incapable of accessing the occluded conformation. ¹H and ¹⁵N CPMG relaxation dispersion analysis for the ternary product complex of the mutant enzyme with NADP⁺ and the product analog 5,10-dideazatetrahydrofolate (ddTHF) (E:ddTHF:NADP⁺) reveals the mechanism by which NADP⁺ is released when the Met20 loop cannot undergo the closed-to-occluded conformational transition. Two excited states were observed, one related to a faster, relatively high-amplitude conformational fluctuation in areas near the active site, associated with the shuttling of the nicotinamide ring of the cofactor out of the active site, and the other to a slower process where ddTHF undergoes small-amplitude motions within the binding site that are consistent with disorder observed in a room-temperature X-ray crystal structure of the N23PP/S148A mutant protein. These motions likely arise due to steric conflict of the pterin ring of ddTHF with the ribose-nicotinamide moiety of NADP⁺ in the closed active site. These studies demonstrate that site-specific kinetic information from relaxation dispersion experiments can provide intimate details of the changes in catalytic mechanism that result from small changes in local amino acid sequence.

Graphical Abstract

Introduction

It is increasingly recognized that biological macromolecules are capable of accessing short-lived higher-energy conformational substates that are critical for their function. Such protein dynamics play essential roles in enzyme catalysis, ligand binding, the molecular basis of allosteric effects, protein folding, and various other processes of biological importance.^1–3 Characterization of weakly populated conformational substates (excited states) sampled through fluctuations on the μs to ms time scale is therefore of utmost importance to understand how evolution has selected specific conformational pathways required for protein function.⁴ Solution NMR spectroscopy is ideal for probing enzyme dynamics at atomic resolution, spanning a wide range of timescales and using many different nuclei.^5–7 Central to the study of exchange processes occurring on the μs-ms time scale are Carr-Purcell-Meiboom-Gill (CPMG) relaxation dispersion (RD) experiments. RD NMR spectroscopy analyzes the effect of slow (µs-ms) stochastic processes on the nuclear spin relaxation, providing valuable information about the kinetics, thermodynamics, and structural changes during the interconversion between the various conformational states involved.^7,8

Escherichia coli dihydrofolate reductase (DHFR) has emerged as a paradigm for understanding the relationship between enzyme structure, dynamics, and catalysis.^2,9–20 DHFR is a small (18kDa) enzyme that catalyzes the NADPH-dependent stereospecific reduction of dihydrofolate (DHF) to tetrahydrofolate (THF), an essential precursor for synthesis of thymine. During catalysis under physiological conditions, E.coli DHFR cycles through five different intermediates (Figure 1A).²¹ Prior studies have shown that the Met20 loop (residues 9–24) switches between two dominant conformations, “closed” and “occluded”, as DHFR progresses through the catalytic cycle.^2,17,22 The enzyme adopts the closed conformation prior to hydride transfer and transitions to an occluded conformation, in which the Met20 loop occludes the binding site for the nicotinamide ring of the cofactor, in the three product complexes (E:THF:NADP⁺, E:THF, and E:THF:NADPH) (Figure 1A). The closed-occluded conformational transition plays a critical role in determining cofactor and product dissociation rates.²³ For the wild-type enzyme, product release is rate-limiting and is facilitated by binding of the NADPH cofactor to the binary E:THF product complex. Conformational fluctuations between the occluded ground state of the product release complex (E:THF:NADPH) and a closed excited state, in which the nicotinamide ring of the NADPH cofactor transiently enters the active site, promote release of THF by an allosteric pathway.^24,25 Mutations that change the relative stability of the closed and occluded states change the kinetics of both product and cofactor release and profoundly modulate the catalytic cycle.^10,15,26,27 By contrast, the Met20 loop of human and other vertebrate DHFRs remains in the closed conformation throughout the catalytic cycle.²⁸

Figure 1.

A. Catalytic cycle of wild type E. coli DHFR. E indicates the DHFR enzyme, DHF dihydrofolate, and THF tetrahydrofolate. B. Catalytic cycle of N23PP/S148A DHFR. Closed complexes are shown in red, occluded complexes in blue. C. Chemical structures of the product tetrahydrofolate and product analog 5,10-dideazatetrahydrofolate. pABG, p-aminobenzoylglutamate. D. Structure of the E. coli DHFR N23PP/S148A dynamic knockout mutant (PDB 3QL0)¹⁰ with bound folate (a substrate analog) and oxidized cofactor NADP⁺.

To delineate the relationship between dynamics and catalysis, Bhabha et al.¹⁰ designed a “dynamic knockout” N23PP/S148A mutant of E. coli DHFR by replacing the E. coli Met20 loop sequence PWN with the PWPP motif from the human enzyme and substituting S148 in the βG-βH loop with Ala to destabilize the occluded conformation. These changes lock the Met20 loop in the closed state, severely impairing hydride transfer, and alter the rate-determining step of the catalytic cycle by inhibiting NADP⁺ dissociation and increasing the rate of product release^10,25 In order to elucidate the molecular mechanism by which the Met20 loop modulates the kinetics of ligand dissociation, we performed NMR relaxation dispersion measurements to probe the dynamics of a complex of the N23PP/S148A mutant with NADP⁺ and a stable product analog, (6R)-5,10-dideazatetrahydrofolate (ddTHF)¹⁷ (Figure 1B). The analog ddTHF was used because the product THF is highly sensitive to light and oxygen. The E:ddTHF:NADP⁺ complex models the product ternary complex of the N23PP/S148A mutant enzyme, for which the rates of THF and NADP⁺ dissociation are increased or decreased ~10-fold, respectively, relative to wild type E. coli DHFR.¹⁰ Our analysis identified two distinct ms time scale processes, one corresponding to the movement of the nicotinamide ring of the cofactor in and out of the active site and the other involving a subset of residues that report on small fluctuations in the product binding site. The current findings provide new insights into the role of the Met20 loop in controlling the kinetics of ligand release and further enhance our understanding of the role of dynamics in enzyme catalysis.

Materials and Methods

General Procedures.

β-nicotinamide adenine dinucleotide phosphate reduced tetrasodium salt hydrate (NADPH), β-nicotinamide adenine dinucleotide phosphate sodium salt hydrate (NADP⁺), folic acid and (6R)-5,10-dideazatetrahydrofolate (ddTHF, also known as lometrexol hydrate) were purchased from Sigma-Aldrich. The N23PP/S148A DHFR was expressed and purified as described elsewhere.^10,29 All experiments were performed in NMR buffer (70 mM potassium phosphate buffer (KPi), 25 mM KCl, 1 mM DTT, pH 7.6) unless otherwise specified.

NMR Spectroscopy.

NMR spectra were acquired at 301K on Bruker Avance spectrometers operating at 500 and 800 MHz. Samples for NMR experiments contained 1 mM ¹⁵N, ²H-labeled N23PP/S148A (∼98% deuteration), 5mM ascorbic acid, 18 mM ddTHF and 10 mM NADP⁺, 1 mM dithiothreitol (DTT), 25 mM KCl, and 10% D₂O in 70 mM KPi pH 7.6. DHFR ligands are extremely sensitive to oxygen and/or light. Buffers were thoroughly degassed through freeze−pump−thaw cycles prior to the addition of ascorbic acid as an oxygen scavenger. All samples were prepared under an argon atmosphere in a glovebox and placed into amber NMR tubes fitted with a VT style valve. Resonance assignments for the E:ddTHF and E:ddTHF:NADP⁺ complexes were made using 3D HNCACB and HN(CO)CACB experiments.^30,31 Relaxation-compensated, constant time CPMG relaxation dispersion data were acquired using published pulse sequences.^32,33 For the N23PP/S148A E:ddTHF:NADP⁺ complex, amide ¹⁵N and ¹H dispersion profiles were recorded at spectrometer frequencies of 500 and 800 MHz using 60% non-uniform sampling (NUS) in the indirect dimension as previously described.^24,34 The total relaxation time for all CPMG experiments was set to 40 ms and the CPMG refocusing frequencies used were 100, 200, 300, 400, 500, 600, 800, 1000, 1200, 1600, 2000 Hz for ¹H CPMG at 800 MHz and ¹⁵N CPMG experiments at 500 MHz and 800MHz. However, for ¹H CPMG experiment at 500MHz, the CPMG frequencies used were 100, 200, 300, 400, 500, 600, 800, 1000, 1200, 1600, 2000, 2400 and 3200 Hz. Errors in peak intensities were estimated using duplicate points collected for the reference CPMG experiment and for the 100, 200 and 2000 Hz data points. Relaxation dispersion data were processed and reconstructed using MDDNMR,^34–36 and NMRPipe.³⁷ Peak intensities were extracted from parabolic fitting using CcpNmr³⁸ analysis and fitted to the Bloch-McConnell equations for two-site exchange using the program Glove.³⁹ Uncertainties in the fitted parameters were estimated by Monte Carlo simulation.

Results and Discussion

Structure of the Ground State.

Chemical shifts of ¹H-¹⁵N HSQC cross peaks for N23PP/S148A E:ddTHF:NADP⁺ are very similar to those of the E:folate:NADP⁺ (Figure S1) and E:THF:NADP⁺ complexes, confirming that it adopts the closed conformation in the ground state.¹⁰ The ¹⁵N chemical shifts of L8, G15, and G95 provide direct evidence that the nicotinamide ring of the cofactor is bound within the active site.⁴⁰ In addition, the characteristically large (5.7 ppm) downfield ¹⁵N chemical shift of A6, relative to the E:folate:NADP⁺ complex, indicates that the pterin ring of the ddTHF occupies the substrate/cofactor binding pocket and, like the actual THF product, forms a hydrogen bond from the N8H to the Ile5 carbonyl.²² The pterin N8 of folate is not protonated and this hydrogen bond cannot form in the folate complex.

The HSQC spectrum of the N23PP/S148A E:ddTHF:NADPH complex is of poorer quality than that of E:ddTHF:NADP⁺ and cross peaks are missing for several residues in the active site and product binding site (Figure S2). Nevertheless, the chemical shifts of critical marker residues (G15, E17, N18, M20, G95, and G121) confirm that the protein adopts the closed conformation with the reduced nicotinamide ring of NADPH occupying the active site.

Relaxation Dispersion Experiments.

In order to evaluate the structural properties of the excited state and the effect of the N23PP/S148A mutations on the μs-ms time scale dynamics of DHFR, we measured amide ¹⁵N and ¹H R₂ relaxation dispersion profiles for the N23PP/S148A E:ddTHF:NADP⁺ complex using relaxation-compensated, constant time CPMG experiments.^32,33 The relaxation dispersion experiments provide quantitative information about μs-ms timescale exchange between conformational states, including the rate of exchange, k_ex, the population (p_B) of the higher energy state(s), and the chemical shift difference Δϖ (in ppm) between the ground and excited state(s). Representative dispersion curves are shown in Figure 2, and the complete set of data for all residues showing dispersion is in Figures S3-S5.

Figure 2.

Representative amide ¹⁵N and ¹H R₂ relaxation dispersion profiles for the E:ddTHF:NADP⁺ complex at 500 MHz (¹⁵N, black; ¹H, red) and 800 MHz (¹⁵N, green; ¹H, blue) spectrometer frequencies. A. Cluster 1. B. Cluster 2.

The dispersion curves of the majority of the residues of the E:ddTHF:NADP⁺ complex could be fitted to a two-site exchange process. Initially, the dispersion data for individual residues were fitted independently. Then, on the basis of the k_ex values obtained from the individual fits, the data were clustered and re-analyzed using a global fitting procedure.

Analysis of the relaxation dispersion profiles of the E:ddTHF:NADP⁺ complex shows that dispersion is associated with residues that are clustered at three distinct locations within the protein. Residues were assigned to the clusters using the criterion χ²_cluster / χ²_individual < 2, where χ²_individual is the reduced χ² value for a given residue when its dispersion profile is fitted independently and χ²_cluster is the reduced χ² value for that residue when fitted to a common k_ex and p_B shared by all residues in the cluster. Residues in the first cluster are mostly located in the Met20 and βF-βG loops and in the vicinity of the NADP⁺ binding site (Figure 3A). The dispersion profiles for these residues fit to a k_ex of 811 ± 11 s⁻¹ and p_B of 4.2 ± 0.1% (Figure S3). The second cluster contains residues located in or connected to the binding pocket for substrate and product (Figure 3B), and the dispersion profiles were fit to a k_ex of 434 ± 9 s⁻¹ and p_B of 13.2 ± 0.5 % (Figure S4). Relaxation parameters obtained from a global fit to the CPMG data for each cluster are shown in Tables S1 and S2. Several residues with weak dispersion profiles could be fitted to either cluster based on the χ² criterion; where possible, these residues were tentatively assigned to cluster 1 or cluster 2 based on their location in the structure, i.e., whether they are spatially contiguous with residues in cluster 1 or cluster 2.

Figure 3.

Structure of the N23PP/S148A E:folate:NADP⁺ complex (3QL0) showing location of residues that exhibit relaxation dispersion in the N23PP/S148A E:ddTHF:NADP⁺ complex. (A) Residues in cluster 1 (k_ex 811 s⁻¹ and p_B 4.2 %) are shown as red spheres. Residues that can be fit to either cluster but are spatially contiguous with cluster 1 are indicated by pink spheres. (B) Residues in cluster 2 (k_ex 434 s⁻¹ and p_B 13.2 %) are shown as blue spheres. Residues that can be fit to either cluster but are spatially contiguous with cluster 2 are indicated by cyan spheres.

The C-terminal associated cluster is composed of residues Y128, E129, D131, D132, and W133 and the best fit of the dispersion profiles was with a k_ex of 670 ± 19 s⁻¹ and p_B 3.2 ± 0.1% (Figure S5). This region exhibits μs-ms timescale conformational fluctuations in all E. coli DHFR complexes studied to date;^29,40–42 these fluctuations are associated with transient formation of an excited state with altered secondary structure of residues 128–133.⁴³

The dispersion profiles for residues 9, 11, and 36 fitted poorly to a two-state exchange process; these residues exhibited three-state behavior arising from a combination of a very fast process and a slow or intermediate-rate process. For other residues such as V10 and R12, the ¹⁵N data fit well to the cluster 1 exchange parameters, and the derived Δϖ_N values correlate well with static chemical shift differences, Δδ (see following section). However, the fit for the ¹H data for these residues is poor, probably because ¹H dispersion is more sensitive to very fast exchange processes than is ¹⁵N dispersion.²⁹ A very fast exchange process, with a k_ex of 4200 s⁻¹, has been observed for active site residues in the E:folate:NADP⁺ complex of N23PP/S148A.²⁹

Structure of the Excited State.

To obtain insights into the structural features of the higher energy excited state(s) sampled by the E:ddTHF:NADP⁺ product complex of N23PP/S148A DHFR, the dynamic chemical shift differences, Δϖ_N and Δϖ_H, between the ground state and excited state were determined by fitting the dispersion curves with the Bloch−McConnell equations. For a subset of residues in cluster 1, located at the beginning of the Met20 loop, in the βF-βG loop, and near the nicotinamide ring of the cofactor, the Δϖ_N and Δϖ_H values determined from the dispersion experiments correlate very closely to the equilibrium chemical shift differences (Δδ_N and Δδ_H) between the E:ddTHF:NADP⁺ and E:ddTHF complexes (Figure 4).

Figure 4.

For cluster 1, correlation between the dynamic chemical shift differences (Δϖ_N,H) obtained from the relaxation dispersion experiments, and the equilibrium chemical shift differences (Δδ) calculated for N23PP/S148A DHFR E:ddTHF:NADP⁺ and E:ddTHF. A.¹⁵N . B.¹H. The diagonal line represents a least-squares fit to the data points.

Unfortunately, equilibrium chemical shift differences (Δδ) are not available for other residues in the Met20 loop (G15, E17, N18, and W22) as the ¹H-¹⁵N cross peaks of these residues are broadened beyond detection in the HSQC spectrum of the binary N23PP/S148A E:ddTHF complex (Figure S6). Residues in cluster 1 that are located in the adenosine phosphate binding site (residues 42, 44, 45, 47, 48, 67, 97, 98, and 102) were excluded from the correlation of Figure 4. The chemical shifts of these residues are dominated by interactions with the cofactor, which remains bound through the adenosine moiety in the excited state (see below). Least-squares fitting of the data plotted in Figure 4 gave a slope of 1.05 and R² of 0.90 for Δϖ_H and a slope of 1.09 and R² of 0.92 for Δϖ_N, indicating a strong correlation between the dynamic and equilibrium chemical shift differences for cluster 1 residues close to the nicotinamide ring of the NADP⁺ cofactor (Figure 3A). The ¹⁵N chemical shifts of L8 and G15 are strongly influenced by hydrogen bonding between the carboxamide moiety of the nicotinamide and the A7 and I14 carbonyl oxygens and are therefore sensitive indicators of nicotinamide occupancy of the active site.^{24,40 15}N chemical shifts are highly sensitive to changes in hydrogen bonding to the carbonyl of the preceding residue, within the same peptide bond.⁴⁴ The values of Δϖ_N for L8 (2.29 ppm) and G15 (2.35 ppm) are similar to the equilibrium chemical shift differences associated with binding of the nicotinamide ring in the active site (Δδ_N ~1.8 and ~2.9 ppm for L8 and G15, respectively, in wild type E. coli DHFR complexes),⁴⁰ providing direct evidence that the exchange process in the closed E:ddTHF:NADP⁺ complex of N23PP/S148A involves transient dissociation of the nicotinamide ring from its binding pocket. The k_ex and p_B values obtained from the cluster fit can be used to estimate the rates of binding (k_in = 777 s⁻¹) and dissociation (k_out = 34 s⁻¹) of the nicotinamide ring from the active site.

The dispersion data for Q102 and H45 show that the adenosine moiety of the cofactor remains bound in the excited state. The side chain of Q102 hydrogen bonds to the adenine ring, resulting in a large downfield shift of a side chain amide cross peak (Figure S1) and confirming that the adenosine is bound in the ground state. The backbone amide proton of H45 forms a hydrogen bond to the phosphate oxygens of the phosphoribosyl moiety. The backbone amide chemical shifts of H45 and Q102 are strongly perturbed by NADP⁺ binding; the equilibrium chemical shift differences between the E:ddTHF:NADP⁺ and E:ddTHF complexes are Δδ_N = 3.37, Δδ_H = 1.22 ppm for H45 and Δδ_N = 1.13, Δδ_H = 0.57 ppm for Q102. The Δϖ values for residues 45 and 102 determined from the dispersion data are much smaller than the equilibrium shift differences (Δϖ_N = 1.05, Δϖ_H = 0.14 ppm for H45, and Δϖ_N = 0.16, Δϖ_H = 0.22 ppm for Q102), showing that the adenosine remains bound in the excited state, even though the nicotinamide ring is dissociated from the active site.

Previous NMR studies of WT E. coli DHFR showed a transition from the closed to the occluded conformation of the Met20 and βF-βG loops on going from the ternary E:folate:NADP⁺ complex to the binary E:THF complex.^2,22,40,45 The chemical shifts of several residues in the active site region have been identified as markers of the closed and occluded conformations.⁴⁶ The equilibrium chemical shift differences (Δδ) for these marker residues between WT E:folate:NADP⁺ (modeling the closed Michaelis complex) and E:THF (the occluded product binary complex) are large and distinctive, whereas the corresponding Δδ values between the E:ddTHF:NADP⁺ and E:ddTHF complexes of N23PP/S148A are much smaller (Figure S7). Although the mutations in N23PP/S148A prevent formation of the occluded structure,¹⁰ the changes in chemical shift differences (Δϖ) between the ground and excited states of E:ddTHF:NADP⁺ and the equilibrium chemical shift differences (Δδ) between the E:ddTHF:NADP⁺ and E:ddTHF spectra show that the Met20 and βF-βG loops can adopt an altered conformation that differs from the closed state. The relaxation dispersion data for the E:ddTHF:NADP⁺ complex provide insights and suggest that the Met20 loop might adopt a partially open conformation in the excited state. An open conformation of the Met20 loop, which appears to be stabilized by crystal packing, has been observed in X-ray structures of E. coli DHFR in some space groups.¹⁷ However, the open state of E. coli DHFR has not been observed by NMR in solution, and therefore no equilibrium Δδ values are available for comparison with the Δϖ values from the dispersion experiments. We therefore predicted chemical shift differences between the open and closed states using the program SPARTA+ and representative crystal structure coordinates. The structures used for the prediction (5Z6F for the closed state, 4X5F (chains A and B) for the open state) were from a pressure-dependent study of E. coli DHFR.⁴⁷ The chemical shift differences for M16, E17, and the amide proton of N18 predicted by SPARTA+ for a transition between the closed to open state are very similar to the Δϖ_N and Δϖ_H values determined by analysis of the relaxation dispersion data (Table S3). In particular, the exceptionally large ¹⁵N chemical shift difference for E17 (Δϖ_N = 6.3 ppm) is in excellent agreement with the predicted difference in chemical shift between the closed and open Met20 loop conformations (Δδ_N = 6.0–6.9 ppm). The mutations introduced in N23PP/S148A would be expected to preclude full opening of the Met20 loop; the hydrogen bonds between N23 and S148 in the βG-βH loop that stabilize both the open and occluded conformations of WT E. coli DHFR cannot form in N23PP/S148A and the C-terminal part of the Met20 loop would be constrained to a conformation similar to that of the closed state. This is consistent with the relatively small Δϖ_N and Δϖ_H values observed for residues 18, 19, 20, an indication that the excited state backbone conformation in this region remains similar to that of the closed ground state. In contrast, large chemical shift differences (Δϖ) would be expected if these residues adopted a fully open conformation in the excited state (Table S3).

Cofactor Release in N23PP/S148A.

The relaxation dispersion data for the faster process (Cluster 1, Figure 3A) provide new insights into the role of conformational fluctuations in release of the cofactor from the closed ground state of the ternary E:ddTHF:NADP⁺ complex of E. coli N23PP/S148A DHFR. The observed correlation between Δϖ and Δδ values for residues near the nicotinamide binding site (Figure 4) and the large Δϖ_N values for L8 and G15 show that the nicotinamide ring exits the active site in the excited state while the adenosine moiety remains bound, as evidenced by the relatively small values of Δϖ for residues in its binding site. A similar motion of the nicotinamide ring is observed in occluded complexes of WT E. coli DHFR,^2,25 where the nicotinamide-out conformation of the occluded ground state transiently samples the nicotinamide-in conformation of the closed complex.

Docking of the nicotinamide in the active site in the closed ground state of the N23PP/S148A E:ddTHF:NADP⁺ complex stabilizes NADP⁺ binding relative to the equivalent WT complex, where the Met20 loop is in the occluded conformation in the ground state and the nicotinamide is dissociated from the binding pocket.⁴⁵ The excited state of the N23PP/S148A E:ddTHF:NADP⁺ complex, in which the nicotinamide ring is released from the pocket while the cofactor remains bound through the adenosine moiety, undoubtedly represents an intermediate on the NADP⁺ dissociation pathway. If we write the dissociation pathway as:

$$\begin{array}{c}\text{E}:\text{ddTHF}:{\text{NADP}}^{+}\\ \text{closed}\end{array}\underset{{\text{k}}_{-1}}{\overset{{\text{k}}_1}{⥋}}\begin{array}{c}\text{E}:\text{ddTHF}:{\text{NADP}}^{+}\\ “\text{open}”\end{array}\stackrel{{\text{k}}_2}{⥋}\text{E}:\text{ddTHF}+{\text{NADP}}^{+}$$

then k_off(NADP⁺) = k₁.k₂/k_-1

For the N23PP/S148A mutant, the rate of dissociation of NADP⁺ from the E:THF:NADP+ complex, k_off(NADP⁺), is 19 s⁻¹;¹⁰ the values of k₁ (34 s⁻¹) and k_-1 (777 s⁻¹) are obtained from fits of the relaxation dispersion profiles of cluster 1 residues (see above). Using these values, we estimate k₂ = 434 s⁻¹ for dissociation of NADP⁺ from the excited state, with the increased dissociation rate reflecting loss of stabilizing interactions with the nicotinamide. This makes perfect sense since the intrinsic rates, measured by stopped flow, for dissociation of NADP⁺ from the occluded THF:NADP⁺ complex of WT DHFR, where the nicotinamide ring is excluded from the active site, are 200 s⁻¹ and higher for some mutants.

Product Release from the N23PP/S148A Mutant.

For Cluster 2, a linear correlation between Δϖ and the Δδ values calculated for (E:ddTHF:NADP⁺ – E:ddTHF) or (E:ddTHF:NADP⁺ – E:NADP⁺) was not observed. The residues in cluster 2 are mostly in or near the substrate/product binding pocket. The Δϖ values for cluster 2 residues are generally small with average values of 0.31 ppm and 0.07 ppm for Δϖ_N and Δϖ_H, respectively. The small size of Δϖ suggests that the exchange processes giving rise to dispersion involve states that are very similar in structure and probably arise from subtle differences in the binding pose of ddTHF. The A6 ¹⁵N resonance is shifted more than 5 ppm downfield in DHF and THF complexes due to the formation of a hydrogen bond between the N8 of the pterin and the CO of I5.^22,25 The large downfield ¹⁵N chemical shift (5.7 ppm) observed for A6 in the E:ddTHF:NADP⁺ complex of N23PP/S148A together with its comparatively small Δϖ_N (0.9 ppm) confirms that the pterin ring of the ddTHF remains in the active site and hydrogen bonded to I5 in the excited state, albeit with altered hydrogen bond geometry. A greater than average value of Δϖ_N (0.89 ppm) is also observed for K58, suggesting changes in the N59 side chain–R57 CO hydrogen bond in the excited state. The N59 side chain forms a network of hydrogen bonds that appear to stabilize the αC-βC loop, which forms one wall of the binding pocket that accommodates the p-aminobenzoylglutamate (pABG) moiety of the ddTHF product analogue. The side chain of R57 binds directly to a pABG carboxyl in an interaction that is buttressed by an R57Nε – P56 CO hydrogen bond. The relaxation dispersion observed for residues in this region provides additional evidence for conformational fluctuations in the product binding site of the N23PP/S148A E:ddTHF:NADP⁺ complex.

The increased rate of THF release from the closed N23PP/S148A E:THF:NADP⁺ complex versus the occluded complex of the WT protein (31 vs 2.4 s⁻¹, respectively)^10,21 likely arises from steric strain in the active site of the mutant enzyme. In the closed conformation, the puckered pterin ring of ddTHF (and by analogy, the THF product) would clash sterically with the nicotinamide ring of the cofactor.¹⁷ In the WT E:ddTHF:NADP⁺ complex, this steric clash is relieved by displacement of the nicotinamide ring from the active site, accompanied by a transition to the occluded conformation. Since the Met20 loop is locked in the closed conformation in the N23PP/S148A mutant, with the nicotinamide ring of NADP⁺ bound in the active site, the steric clash between the nicotinamide and pterin must be relieved in some other way. Room temperature X-ray structures of the closed E:folate:NADP⁺ complexes of WT E. coli DHFR and the N23PP/S148A mutant revealed increased flexibility and conformational heterogeneity in the active site of the mutant enzyme that propagates across the β-sheet.⁴⁸ The increased active-site flexibility in N23PP/S148A leads to disorder in the positions of both the NADP⁺ and folate, with two binding poses of the pterin ring that differ in the I5 CO-N8 distance. Based on the X-ray structure, it appears likely that N23PP/S148A DHFR can accommodate both the nicotinamide ring and the puckered pterin ring of THF within the closed active site with reduced steric clash relative to WT DHFR. It is likely that the increased conformational heterogeneity of the N23PP/S148A mutant is manifest in the relaxation dispersion behavior of residues in cluster 2, which undergo conformational exchange at a rate of 434 s⁻¹ and with a minor state population of 13.2%. Many of these residues, which are located in or near the substrate/product binding pocket and across the β-sheet (Figure 3B), exhibit conformational disorder in the room-temperature X-ray structure of the N23PP/S148A E:folate:NADP⁺ complex.⁴⁸ The Δϖ values for residues in cluster 2 are generally small, showing that there are only very small conformational changes between ground and excited states, consistent with the local heterogeneity observed in the room temperature X-ray structure.⁴⁸ It is not clear whether transitions into the excited state are functional in promoting product release or whether they simply reflect flexibility and conformational heterogeneity in the mutant enzyme. The relaxation dispersion data for A6 shows that the pterin ring of ddTHF is bound in the active site, with its N8 atom hydrogen bonded to the I5 CO, in both the ground and excited states. Changes in the A6 ¹⁵N chemical shift between ground and excited states suggest perturbation of the hydrogen bond geometry in the excited state, but whether this is an intermediate on the product dissociation pathway remains an unanswered question.

For WT E. coli DHFR, the rate determining step in the catalytic cycle is THF dissociation from the occluded ternary product release complex, E:THF:NADPH.²¹ THF release is facilitated by an allosteric mechanism, in which the nicotinamide ring of the cofactor transiently enters the active site to form a closed excited state; the resulting steric clash between the puckered nicotinamide ring of the reduced cofactor and the puckered pterin ring of the THF accelerates product release.^24,25 In contrast, the E:THF:NADPH complex of N23PP/S148A is already in the closed conformation in its ground state, with the nicotinamide bound in the active site, and the resulting steric strain increases the intrinsic rate of THF dissociation from 3.7 s⁻¹ for WT DHFR to 51 s⁻¹ for the mutant enzyme.²⁴

Alternative Catalytic Pathway for N23PP/S148A.

By locking the Met20 loop into the closed conformation, the N23PP/S148A mutation alters the preferred catalytic pathway of the enzyme. Because the ground state of the ternary complex E:ddTHF:NADP⁺ remains in the closed conformational state, the rate of dissociation of the oxidized cofactor is decreased by a factor of 10, from 200 s⁻¹ in the WT protein to 19 s⁻¹ in the mutant protein.^10,21 At the same time, the dissociation rate of the reduced product THF (modeled in our studies by the stable analog ddTHF) is increased by a factor of 10, from 2.4 s⁻¹ in the WT protein to 31 s⁻¹ in the mutant protein. These changes have the effect of promoting ligand flux through an alternative pathway where the product can dissociate before the canonical steps in the WT pathway associated with release of oxidized cofactor and rebinding of reduced cofactor (Figure 5). There is also a change in the rate limiting step of the catalytic cycle, from THF dissociation from the E:THF:NADPH complex for the WT enzyme to NADP⁺ dissociation from E:NADP⁺ for N23PP/S148A.

Figure 5.

Catalytic cycle for E. coli DHFR, showing the canonical cycle for the WT enzyme (black) and the alternate pathway enabled by the N23PP/S148 mutations. Rate constants for the dissociation of product and oxidized cofactor are shown in black for WT and red for the N23PP/S148 mutant in complex with ddTHF and NADP⁺.

Conclusions

The present work provides new insights into the role of the Met20 loop in modulating ligand flux in E. coli DHFR. Substitution of N23 of the bacterial enzyme with the PWPP motif from human DHFR reduces the flexibility of the Met20 loop and inhibits the closed to occluded transition in the product complexes while the S148A substitution further destabilizes the occluded conformation.^10,23 In combination, these mutations strongly increase the affinity for cofactor, decreasing the rate of dissociation of NADP⁺ from the E:THF:NADP⁺ complex by 10-fold.¹⁰ Millisecond time scale conformational fluctuations in the Met20 and βF-βG loops play a critical role in controlling the flux of ligands in E. coli DHFR. The present work shows that such fluctuations facilitate cofactor release from the closed ground state of N23PP/S148A via a transient intermediate, possibly with a partially open loop conformation, in which the nicotinamide ring is displaced from its binding pocket and the cofactor remains bound only through its adenosine moiety. In contrast, fluctuations of the Met20 loop in the occluded ground state of the wild type DHFR result in transient entry of the nicotinamide ring into the active site and promote dissociation of product by an allosteric mechanism.^24,25 The dynamics of the Met20 loop appear to play no direct role in determining the kinetics of product release from N23PP/S148A, which is more than 10-fold faster than from wild type DHFR.¹⁰ Rather, product binding is destabilized by steric clash between the pterin and nicotinamide rings in the closed ground state of the mutant enzyme, leading to millisecond time scale exchange between alternate binding modes that may facilitate product dissociation. In contrast, product is bound tightly to the occluded ground state of the wild type E. coli DHFR; release of product is facilitated by an allosteric mechanism in which fluctuations of the Met20 loop result in transient entry of the nicotinamide ring into the active site and steric clash with the pterin ring.^24,25 Thus, Met20 loop dynamics play very different roles in the occluded product complexes of the wild-type enzyme and in the closed product complexes of the N23PP/S148A mutant, promoting product release in the former and cofactor release in the latter.

The emergence of a conformationally restricted proline-rich Met20 loop is a late evolutionary adaptation that allows human and other vertebrate DHFRs to bind cofactor tightly and thus function at the very low concentrations of NADP⁺ in eukaryotic cells.⁴⁹ In contrast, E. coli DHFR must function at 100-fold higher cellular NADP⁺ concentration; the enzyme has evolved a dynamic catalytic mechanism in which the Met20 loop undergoes large-scale conformational changes to occlude the nicotinamide binding site, thereby promoting cofactor release and avoiding product inhibition.