Comparison of the Role of Protein Dynamics in Catalysis by Dihydrofolate Reductase from E. coli and H. sapiens

Abstract

Dihydrofolate reductase (DHFR) is a well-studied, clinically relevant enzyme known for being highly dynamic over the course of its catalytic cycle. However, the role dynamic motions play in the explicit hydride transfer from the nicotinamide cofactor to the dihydrofolate substrate remains unclear because reaction initiation and direct spectroscopic examination on the appropriate time scale for such femtosecond to picosecond motions is challenging. Here, we employ pre-steady-state kinetics to observe the hydride transfer as directly as possible in two different species of DHFR: Escherichia coli and Homo sapiens. While the hydride transfer has been well-characterized in DHFR from E. coli, improvements in time resolution now allow for sub-millisecond dead times for stopped-flow spectroscopy, which reveals that the maximum rate is indeed faster than previously recorded. The rate in the human enzyme, previously only estimated, is also able to be directly observed using cutting-edge stopped-flow instrumentation. In addition to the pH dependence of the hydride transfer rates for both enzymes, we examine the primary H/D kinetic isotope effect to reveal a temperature dependence in the human enzyme that is absent from the E. coli counterpart. This dependence, which appears above a temperature of 15 °C is a shared feature among other hydride transfer enzymes and is also consistent with computational work suggesting the presence of a fast promoting-vibration that provides donor–acceptor compression on the time scale of catalysis to facilitate the chemistry step.

Graphical Abstract

INTRODUCTION

Dihydrofolate reductase (DHFR) has served as a paradigm for the role of protein dynamics in enzyme catalysis. It is a biologically ubiquitous enzyme that utilizes a nicotinamide adenine dinucleotide phosphate (NADPH) cofactor to reduce dihydrofolate to tetrahydrofolate via a hydride transfer and coupled proton transfer. Tetrahydrofolate is a crucial intermediate in multiple biosynthetic pathways including the generation of nucleic acids, making DHFR an important metabolic enzyme.^1–6 Human DHFR is the enzymatic target for the commercially available chemotherapy agent methotrexate;⁷ other drugs designed to halt bacterial⁸ or parasitic⁹ DNA synthesis also target DHFR. Well-characterized by X-ray crystallography¹ and nuclear magnetic resonance (NMR) spectroscopy,^10–12 DHFR is known to undergo conformational changes associated with enzymatic turnover. The multifaceted clinical relevance, dynamic nature of the catalytic cycle, and the hydrogen transfer processes catalyzed all serve to make DHFR an interesting subject for further study, including fundamental investigations into the mechanism(s) of rate acceleration in enzymes.

A thorough investigation of the kinetics of the catalytic cycle in DHFR from Escherichia coli, including a complete kinetic scheme was published by Fierke, Johnson, and Benkovic in 1987.¹³ Using stopped-flow fluorescence, the authors measured the pre-steady-state hydride transfer rate across a range of pH values and established a maximum rate at low pH of 950 s⁻¹ with a maximum primary H/D kinetic isotope effect (KIE) of 3. DHFR was examined as a candidate for containing a rate promoting vibration because of its H/D KIE and the possibility of a contribution from quantum mechanical tunneling to the hydride transfer process.^14–16 Rate promoting vibrations are a class of enzyme dynamics that occur on the time scale of catalysis (femtosecond to picosecond) and provide an oscillatory compression that modulates the hydrogen donor–acceptor distance, and thus, the hydride transfer efficiency. Using a transition path sampling method, the Schwartz group had already identified such motions in another hydrogen transfer enzyme, human heart lactate dehydrogenase.¹⁷ While dynamics on the time scale of catalysis were observed in E. coli DHFR, the motions did not appear to correlate to a compression along the reaction coordinate.¹⁸

Wright and co-workers performed a phylogenic analysis of DHFR from E. coli to Homo sapiens and through NMR, X-ray crystallography, and bioinformatics established that while the species are structurally very similar, their dynamics differ.¹⁹ The human enzyme, despite being larger, is less flexible and undergoes more exaggerated hinge-type motions. An inspection of the two enzyme structures also revealed that the human enzyme had multiple major coordinated evolutionary mutations that reshaped the active site, creating the possibility of a linear axis through the NADPH-DHF donor–acceptor axis. Schwartz and Masterson repeated their computational promoting vibration analysis for DHFR from H. sapiens and found that a compressive promoting vibration was indeed present, suggesting the motion was evolutionarily designed.²⁰

Low frequency (50–200 cm⁻¹) promoting vibrations in enzymes have not yet been observed directly, due to technical challenges. An alternative approach, developed by Schramm, is the “Born–Oppenheimer Enzyme,” where all nonexchangeable carbon, nitrogen, and hydrogen atoms have been exchanged for their heavier isotopes, effectively producing a “heavy” enzyme with ~10% greater mass, in order to decouple the native enzyme dynamics from catalysis.²¹ For human purine nucleoside phosphorylase, the heavy enzyme exhibits reduced single-turnover rates, suggesting that the femtosecond dynamics in the “light” enzyme contribute to enzymatic catalysis. A similar approach was taken with heavy DHFR from E. coli and no alteration to the hydride transfer rate was observed;^14,16 this is in agreement with the previous computational work claiming the enzyme from E. coli lacks a promoting vibration. Allemann and co-workers generated a heavy ecDHFR where the hydride transfer was observed to be slower than the nonisotopically substituted light counterpart, but the absence of a difference in barrier heights and tunneling contributions suggested to the authors that the difference was due to coupling of global protein motions not specifically contributing to the hydride transfer.²²

A “heavy” equivalent of human DHFR was also produced,²³ but the isotopic substitution altered the steady-state kinetics of the enzyme, a phenomenon not observed with other heavy enzyme counterparts. This observation suggested that more than the catalytic step of interest, the hydride transfer, was affected by the change in mass. The pre-steady-state single turnover kinetics for the human enzyme were experimentally inaccessible due to the speed of the reaction; the hydride transfer was unable to be monitored via stopped-flow fluorescence as the decay was lost to the 2 ms dead time of the instrument.

In this study, we employ a stopped-flow instrument with a greatly improved, sub-millisecond dead time capable of directly observing the pre-steady-state hydride transfer in human DHFR for the first time. In addition, we revisit the original kinetic scheme of E. coli DHFR and challenge the maximum rate of hydride transfer at low pH. To determine the dynamic effects present or absent in each species, we employ temperature dependent primary kinetic isotope effects, where a relationship of the KIE with temperature indicates a nonuniform barrier modulated by the internal dynamics coupled to catalysis.^24,25 We are able to establish a new lower limit on the maximum hydride transfer rate of DHFR for both species and provide the first experimental evidence for fast, femtosecond dynamics coupled to catalysis in the human enzyme.

MATERIALS AND METHODS

Unless otherwise specified, all materials were obtained from Sigma-Aldrich and were of the highest purity available. All solutions were prepared fresh daily from solids.

Expression and Purification of His-Tagged E. coli and H. sapiens DHFR.

Wildtype DHFR from E. coli containing a C-terminal six-histidine tag following a TEV-protease cleavage site was cloned and expressed in E. coli strain BL21 (DE3) with Luria–Bertani (LB) medium and 100 μg/mL ampicillin. A single colony was isolated and inoculated into 20 mL of LB medium at 30 °C overnight, after which 1 mL of this starter colony was inoculated into 1 L of additional medium at 37 °C and the bacterial growth was monitored until OD₆₀₀ = 0.6–0.8. Isopropyl β-d-thiogalactopyranoside (IPTG) was added (1 mM final concentration) and the culture was incubated overnight at 30 °C with shaking at 200 rpm. Cells were harvested via centrifugation (5000 g/15 min/4 °C) and stored at −80 °C until purification. Thawed cell pellets were resuspended in 50 mM Tris pH 8.0, 150 mM NaCl, 5 mM β-mercaptoethanol, 1 mg/mL lysosome, and 1 tablet of protease inhibitor per 50 mL of lysis buffer. After stirring on ice for 1 h, the cells were sonicated on ice (Sonic Dissemble model 500, Fisher Scientific, Pittsburgh, PA) and insoluble debris was precipitated via centrifugation (13000 rpm/30 min/4 °C). The supernatant was filtered through a 0.22 μm filter and purified using a HisPrep affinity column on an AKTA FPLC system (GE Healthcare, Pittsburgh, PA). Prior to application of the DHFR lysate the column was equilibrated with 50 mM TrisHCl pH 8.0, 150 mM NaCl, 5 mM β-mercaptoethanol, and 10 mM imidazole. Protein elution was performed via a gradient over 15 column volumes from 100% equilibration buffer to 75% equilibration buffer and 25% of an elution buffer composed of 50 mM Tris-HCl pH 8.0, 150 mM NaCl, 5 mM β-mercaptoethanol, and 500 mM imidazole. Eluted protein was concentrated using a 10 kDa molecular weight cutoff filter in an Amicon centrifugal concentrator and then exchanged via a HiPrep desalting column into a storage buffer containing 50 mM sodium phosphate pH 7.0, 100 mM NaCl, 5 mM DTT, and 5% glycerol. Protein was stored at −80 °C until cleavage and use in kinetic experiments. Enzyme purity and molecular weight was validated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and protein concentration was quantified using a molar extinction coefficient of ε_{280 nm} = 31,300 M⁻¹ cm⁻¹. A plasmid containing his-tagged human DHFR with a TEV cleavage site (pET21a-His-TEV-hDHFR) was a gift from the lab of Andrew Lee (UNC Chapel Hill, Eschelman School of Pharmacy). Expression and purification followed the same general procedure as that of E. coli DHFR but in addition incorporated 1 mM folic acid in the buffer used for resuspending thawed cell pellets and in all subsequent buffers. The extinction coefficient used for the quantification of the human enzyme was 25,440 M⁻¹ cm⁻¹, but due to the excess folic acid and oxidized cofactor present, it is inadvisible to use absorbance to quantify the protein concentration prior to further purification.

Proteolytic TEV-Site Cleavage.

The plasmid for histidine-tagged recombinant TEV protease was a gift from the lab of Emily Weinert (Penn State University), and the protease was expressed in E. coli. His-tagged DHFR was cleaved using a protocol²⁶ from the Wedekind laboratory at the University of Rochester Medical Center. Briefly, frozen DHFR stocks in storage buffer (up to 30 mg of enzyme) were thawed at 4 °C and diluted to approximately 2 mg/mL in a low salt phosphate buffer (100 mM sodium phosphate, pH 7.0, no added NaCl). TEV protease was added at a ratio of 1:50–1:100 w/w TEV protease:DHFR, pipetted gently to mix, and left undisturbed overnight to react at 4 °C. Uncleaved DHFR, the his-tagged TEV protease, and the free His-tag sequences were collected via affinity chromatography on a gravity column filled with NiNTA HisPur Resin (Thermo), and the flow-through was pooled and concentrated via centrifugation (Amicon 10 kDa MWCO 15 mL insert). Recovery of E. coli DHFR typically exceeded 95% following cleavage, while human DHFR recovery was generally closer to ~65%.

Dihydrofolate Substrate Synthesis.

The dihydrofolate used as a substrate for all kinetic experiments was synthesized via reduction of folic acid with dithionite.²⁷ Briefly, 20 g of ascorbic acid was stirred to dissolve in 100 mL deionized water and the solution was adjusted to pH 6.0 using 1 M NaOH. A solution of folic acid dihydrate (800 mg, 1.8 mmol in 5 mL of deionized water) was added to the ascorbic acid solution, which was then cooled on ice. Once the solution temperature reached 4 °C, solid dithionite (15 g, 86 mmol) was added and the solution was allowed to continue stirring on ice for an additional 20 min. Then 6 M hydrochloric acid was added dropwise (~15 mL) to acidify the solution to pH = 2.8, after which the solution was centrifuged for 15 min at 4750 rpm and 4 °C. Following centrifugation, the supernatant was discarded, and the pellet was resuspended in 200 mL sodium ascorbate buffer, pH 6. The solution was cooled to 4 °C, reacidified using 6 M HCl to pH = 2.8, and kept on ice for an additional 10 min to precipitate the solid. Following a second round of centrifugation (15 min, 4 °C, 4750 rpm), the wet orange solid was suspended in 1 mM HCl and transferred to tared falcon tubes for flash freezing in liquid nitrogen and lyophilization to remove the water. Following lyophilization, yield by mass was determined to be 94% (0.757 g, 1.7 mmol) and the product was characterized via UV–vis (A_max = 282 nm). Synthesized DHF was divided into aliquots and stored in amber vials under nitrogen at −20 °C.

Synthesis and Purification of R [4-²H]-NADPH.

Stereo-specific isotopic labeling of NADPH followed the enzymatic procedure published by Jeong and Gready.²⁸ Alcohol dehydrogenase from Thermoanaerobacter brocki (3.78 U, Sigma-Aldrich, recombinant from E. coli, > 500 U/mL), 35 mg NADP⁺ disodium salt (2.6 mM), 1.2 mL isopropanol-d⁸ (1 M, Cambridge Isotope Laboratories) and 15 mL 25 mM TRIS pH 9.0 were combined in a 20 mL glass scintillation vial. The vial was stirred/shaken gently on a hot plate set at 43 °C and reaction progress was monitored every few minutes by drawing 5 μL aliquots and measuring the solution absorbance at 340 nm on a Nanodrop spectrometer (Thermo Fisher Scientific). A maximum A₃₄₀ was achieved after about an hour, reaching 80% conversion to NADPH (ε₃₄₀ = 6,220 M⁻¹ cm⁻¹). The vial was placed on a rotary evaporator to remove the acetone product and any unreacted ethanol before the remaining aqueous solution was filtered via centrifugation (10 kDa MWCO Amicon, 15 mL) to remove the enzyme. 100 mg NaCl was added to the filtrate which was then flash frozen in liquid nitrogen and lyophilized while protected from light. A reverse-phase preparative scale HPLC method for the purification of the lyophilized solid was adapted from that of Markham, Sikorski, and Kohen.²⁹ A preparative C18 column (25 mm × 250 mm, 5 μm) was equilibrated with 200 mM NaCl in 1 mM TRIS pH 8.2 (Buffer A) at 14 mL/min for an hour before injecting the crude material. Following injection, a gradient was followed as in Table 1 to 100% MeOH.

Table 1.

NADPH purification by preparative HPLC on a C18 Column²⁸

time (min)	% buffer Aa	% MeOH
0.0	100	0
11.0	100	0
11.1	85	15
21.0	85	15
23.0	0	100
27.0	0	100

^aBuffer A is 200 mM NaCl in 1 mM TRIS pH 8.2.

Elution was monitored via absorbance at both 260 and 340 nm, and fractions containing A_260/340 < 2.5 were collected (retention time ~ 17 min). The pooled fractions were spiked with 1% v/v 1 M TRIS-HCl pH 9.2 prior to flash freezing in liquid nitrogen and lyophilization, protected from light. The dry solid was stored at −80 °C, and the A_260/340 was ensured to be <0.25 prior to use.

Kinetic Assays.

DHFR was purified as described above and prior to kinetic experiments the protein was exchanged into MTEN buffer (50 mM 2-(N-morpholino)ethane-sulfonic acid (MES), 25 mM tris(hydroxymethyl)aminomethane, 25 mM ethanolamine, and 100 mM sodium chloride) at the indicated pH. DHFR and solutions of DHF and NADPH(D) were all quantified spectrophotometrically using extinction coefficients of ε₂₈₀ = 31,100 M⁻¹ cm⁻¹ (E. coli DHFR), ε₂₈₀ = 25,440 M⁻¹ cm⁻¹ (H. sapiens DHFR), ε₂₈₂ = 28,000 M⁻¹ cm⁻¹ (DHF). And ε₃₄₀ = 6,200 M⁻¹ cm⁻¹ (NADPH(D)). Final concentrations, unless otherwise specified, were 15 μM DHFR, 125 μM NADPH, and 100 μM DHF. Single turnover results are the average of at least 15 consecutive shots, and the error for each reported rate is the standard deviation of multiple independent trials, and in the case of human DHFR, multiple expression batches. All data were fit analytically using Igor Pro, Version 7.0 (Wavemetrics). The pre-steady-state hydride transfer was monitored as previously described via Förster Resonance Energy Transfer (FRET) fluorescence, where excitation of the native tryptophan residues in DHFR at 280 nm results in FRET to the bound NADPH cofactor, which then emits at 460 nm. The reaction is initiated by the rapid mixing of the enzyme:cofactor complex with the substrate, and monitored by the decrease in the emission signal at 460 nm caused by hydride transfer and oxidation of the NADPH cofactor. Kinetic measurements were collected on a commercially available 2-syringe BioLogic Stopped-Flow mixing system (dead time 313 μs at a total flow rate of 14 mL/s, valve lead 2.6 ms) equipped with a National Instruments data acquisition card for digital data collection at a sampling rate of 10 μs. Excitation light was provided via a fiber optic connection from a Xe lamp excitation source (Mikropak HPX-2000) equipped with a 290 ± 5 nm bandpass filter (Edmund Optics) and fluorescence emission intensity was detected via a photomultiplier tube detector (Thor Laboratories PMM02, external DC voltage power supply from TackLife) with a 460 ± 30 nm bandpass filter. The temperature during all measurements was maintained by a recirculating water bath (Thermo, RTE7) at 25 °C unless otherwise specified. Further instrumental details, including experimental validation of the dead time are included in the Supporting Information.

RESULTS AND DISCUSSION

pH Dependence of the Hydride Transfer in E. coli DHFR.

The pH dependence of the hydride transfer in wildtype DHFR from E. coli was determined using stopped-flow fluorescence. Representative pre-steady-state kinetic traces measured by FRET from Trp to bound NADPH across a range of pH values from 5.0 to 9.0 are presented in Figure 1, clearly showing the decrease in the exponential rate with increasing solution pH. The hydride transfer rate at each pH value is included in Table 2. When the hydride transfer rate is plotted as a function of pH, the expected sigmoidal curve is generated. DHF reduction requires both a hydride transfer and a proton transfer, thus at high pH (low [H⁺]), proton transfer is assumed to be rate-limiting and the kinetics report on the proton transfer rate, while the low pH (high [H⁺]) regime provides the rate of the explicit hydride transfer as the substrate is already protonated and the rate is not delayed by proton transfer. The transition between the two regimes represents the Asp27 side chain and coupled substrate protonation equilibrium and fitting the rates as a function of pH yields the pK_a of this residue in the substrate- and cofactor-bound complex.

Figure 1.

pH dependence of hydride transfer rate for WT ecDHFR at 25 °C. (A) Kinetic traces measured by FRET across the range of pH values included in the study, demonstrating the inverse relationship of hydride transfer rate with pH. Single exponential fits are shown in black. (B) Hydride transfer rate as a function of pH, fit to a Henderson–Hasselbalch equation. Fit predicts pK_a = 6.2 ± 0.1, maximum rate = 1100 ± 47 s⁻¹. Error bars represent standard deviation of the rate from at least 2 independent trials, each consisting of at least 15 consecutive shots.

Table 2.

Observed Hydride Transfer Rates for WT ecDHFR^a

average pH (all values ±0.02)	average rate ± std dev (s−1)
5.00	1096 ± 2
5.25	951 ± 38
5.50	809 ± 20
6.00	588 ± 24
6.50	363 ± 18
6.70	282 ± 16
7.05	170 ± 1
7.68	81 ± 1
8.10	50 ± 4
8.56	35 ± 1
8.99	25 ± 1

^aError represents standard deviation between n > 2 trials of at least 15 averaged stopped flow traces per pH value tested.

Figure 1B displays the observed hydride transfer rate as a function of pH fit to a rearranged Henderson–Hasselbalch equation:

$$\text{rate}=\text{maximum\hspace{0.17em}rate\hspace{0.17em}}\times \frac{1}{{10}^{\text{pH}-p{\text{K}}_{\text{a}}}+1}$$ (1)

where the rates represent the fractional protonation relative to a maximum rate.

Notably, the measured rate at low pH exceeds the previously determined maximum of 950 s⁻¹;¹³ we observe k_hyd = 1096 ± 2 s⁻¹ for pH 5.0. The maximum rate predicted by the Henderson–Hasselbalch equation is approximately 1100 s⁻¹, in excess of the previously determined rate employed in the kinetic scheme of ecDHFR of 950 s⁻¹. The increased rates for the present study are attributed to the improved time resolution (decreased dead time from 1.6 ms to sub-millisecond) and thus the ability to directly observe nearly the full exponential decay in fluorescence. The hydride transfer rates for pH 6.0 and higher are very consistent with the original values observed¹³ and a sigmoidal fit excluding any rates faster than 700 s⁻¹ (faster than the dead time of the instrument used previously) faithfully reproduces the previous curve and predicts an upper limit of 1000 ± 92 s⁻¹.

As a result of the increased rates at lower pH, the pK_a for the hydride transfer is also shifted; originally determined to be 6.5, our data places the pK_a at a slightly more acidic 6.2 ± 0.1. This pH dependence was originally attributed to the active site Asp27 of the ternary complex with reduced cofactor and substrate bound, DHFR:NADPH:DHF.¹³ More recently, the apparent pK_a of the hydride transfer was attributed to protonation of the N5 on the pterin ring of the nictotinamide cofactor, because Asp27 was shown to remain unprotonated throughout the reaction cycle by Raman spectroscopy.^5,31–33 If the pH dependence of the rate represents a strictly monoprotic acid–base equilibrium (i.e., protonation of N5), however, we would expect a better fit to the HH equation. The observed deviation from HH behavior is evidence for a more complex process, such as a proton transfer pathway with multiple (coupled) protonatable groups. This observation is consistent with previous proposals for a hydrogen-bonded network in DHFR that delivers protons to the substrate. It is also possible that enzyme samples multiple kinetically competent populations, resulting in small deviations from the expected HH behavior.^5,30–33

Temperature Dependence of the Primary H/D KIE for the Hydride Transfer in E. coli DHFR.

Temperature dependent kinetic isotope effects have been proposed as experimentally observable evidence for the presence of promoting vibrations.^24,25 In order to perform a direct comparison between the two species of DHFR, the temperature dependence of the primary kinetic isotope effect for the hydride transfer was studied using deuterated NADPH. The kinetics were collected at high pH (pH 9.0) for the sake of comparison to the previously published KIE data for the E. coli enzyme. As the hydride transfer at high pH is quite slow, no difference from the previously reported KIE value of approximately 3 was anticipated. The kinetic traces collected across temperatures from 5 to 45 °C are shown in Figure 2. Rates accompanying each temperature are enumerated in Table 3 and the relationship between the determined KIE and temperature is displayed in Figure 2F.

Figure 2.

Kinetic traces for the ecDHFR hydride transfer used to calculate kinetic isotope effects. Traces where NADPH was used as a cofactor are shown in light blue, and traces following trials with NADPD are shown in dark blue. Each trace represents the average of at least 15 consecutive stopped flow shots and is fit to single exponential decay, shown in red. Temperatures surveyed include (A) 5 °C, (B) 15 °C, (C) 25 °C, (D) 35 °C, and (E) 45 °C. (F) Temperature dependence of hydride transfer rate KIE for ecWTDHFR.

Table 3.

ecWTDHFR Hydride Transfer Rates Used to Determine Primary KIE across Temperatures 5–45 °C

temperature (°C)	hydride transfer rate with light isotope, kNADPH (s−1)	hydride transfer rate with heavy isotope, kNADPD (s−1)	KIE (kNADPH/kNADPD)
5	21.3	7.07	3.02
15	31.8	9.95	3.19
25	30.0	10.1	2.96
35	38.8	13.1	2.96
45	45.7	16.2	2.82

The temperature independent KIE for the ecDHFR is in agreement with previously published work and the predicted computational evaluations that find no evidence of a promoting vibration.^14,16,17 The barrier to hydride or deuteride transfer appears uniform throughout the range of temperatures included in this study.

pH Dependence of the Hydride Transfer in H. sapiens DHFR.

In contrast to the bacterial enzyme, the more evolved human enzyme is thought to have a remodeled active site and an axis of residues that constitute a promoting vibration.²⁰ Until the present study, a thorough analysis of the pH dependence and the KIE of the hydride transfer in human DHFR using pre-steady-state kinetics has not been attempted due to the anticipated fast rate and multi-millisecond dead times of most commercial stopped-flow instruments.^4,22

With an improved, sub-millisecond dead time for stopped-flow fluorescence, we were able to measure directly the pre-steady-state hydride transfer in DHFR from H. sapiens for the first time. The kinetic traces collected from pH 9.0 to 7.0 are presented in Figure 3A. At 25 °C, the hydride transfer rate for DHFR from E. coli is 170 s⁻¹, but for the human enzyme, the rate is approximately 10-fold faster, k_hyd = 1898 s⁻¹. Hydride transfer rates at each pH are listed in Table 4 and plotted in Figure 3C. An attempt was made to quantify the hydride transfer rate at pH 6.5 via FRET, but an average of 19 consecutive stopped-flow shots yielded a transient with low signal-to-noise. Monitoring the direct NADPH excitation instead (λ_ex = 360 nm λ_em = 460 nm), the trace presented in Figure 3B was obtained, with a single exponential fit of 3263 s⁻¹, which is close to the limit of what the instrument is able to resolve (dead time = 313 μs). When all observed rates, including pH 6.5 are fit to the Henderson–Hasselbeck equation, the pK_a for the Asp27 side chain is predicted to be 6.9 and the maximum rate at low pH is estimated at 4600 s⁻¹. As the low pH data may be instrument-limited, this analysis likely underestimates the maximum hydride transfer rate and overestimates the pK_a.

Figure 3.

pH dependence of hydride transfer rate for hsWTDHFR at 25 °C. (A) Kinetic traces measured by FRET for pH 7.00 (red-orange), 7.50 (yellow-orange), 8.00 (green), 8.50 (blue), and 9.00 (purple) demonstrate the inverse relationship of hydride transfer rate with pH. Each trace represents an average of at least 14 consecutive stopped flow shots, with single exponential fits shown in black. (B) Kinetic trace for hsWTDHFR hydride transfer at pH 6.5 measured by direct excitation of NADPH (23 shot average); single exponential fit in black reveals a rate of 3263 s⁻¹, approaching the instrumental dead time. (C) Hydride transfer rate as a function of pH, fit to the Henderson–Hasselbalch equation; pK_a = 6.9 ± 0.05, maximum rate = 4600 ± 240 s⁻¹.

Table 4.

Observed Hydride Transfer Rates for hsWTDHFR^a

average pH (all values ±0.02)	average rate ± std dev (s−1)
6.50	3263 ± 543
7.00	1898 ± 79
7.50	989 ± 53
8.00	272 ± 23
8.50	54 ± 6
9.00	25 ± 4

^aError represents standard deviation between n > 2 trials of at least 15 averaged stopped flow traces per pH value tested.

Temperature Dependence of the Primary Kinetic Isotope Effect in H. sapiens DHFR.

As the stopped-flow method was clearly able to resolve the hydride transfer rate for the human enzyme at high pH, we were able to determine the primary KIE for the hydride transfer for the first time. Kinetics were collected using NADPH and the deuterated cofactor NADPD across a range of temperatures, show in in Figure 4.

Figure 4.

Kinetic traces for the hsDHFR hydride transfer used to calculate kinetic isotope effects. Traces where NADPH was used as a cofactor are shown in light blue, and traces following trials with NADPD are shown in dark blue. Each trace represents the average of at least 15 consecutive stopped flow shots and is fit to single exponential decay, shown in red. Temperatures surveyed include (A) 5 °C, (B) 15 °C, (C) 25 °C, (D) 35 °C, and (E) 45 °C. (F) Temperature dependence of hydride transfer rate KIE for hsWTDHFR.

Examination of the KIE at low temperature revealed the expected value of 3, consistent with the results from the E. coli enzyme. However, as the temperature increased, the difference in rates between the cofactor containing the heavy and light isotopes was inconsistent; as shown in Table 5 and Figure 4, with increasing temperature the KIE appears to decrease.

Table 5.

hsWTDHFR Hydride Transfer Rates Used to Determine Primary KIE across Temperatures 5–45 °C^a

temperature (°C)	Hydride transfer rate with light isotope, kNADPH (s−1)	hydride transfer rate with heavy isotope, kNADPD (s−1)	KIE (kNADPH/kNADPD)
5	7.5 ± 3	2.5 ± 1	2.9 ± 0.05
15	12±2	4.3 ± 0.9	2.9 ± 0.2
25	25 ± 0.3	15±2	1.6 ± 0.2
35	54 ± 9	37 ± 13	1.5 ± 0.3
45	88 ± 6	66 ± 15	1.4 ± 0.2

^aValues shown are averages of 2 trials across two expression batches ± one standard deviation.

This temperature dependence of the primary KIE indicates that the relative transfer rate for a hydride vs deuteride is temperature dependent, suggesting a role for temperature dependent protein motions in modulating the barrier to hydride/deuteride transfer. Increased temperatures yielded smaller KIE values, possibly due to a smaller impact from tunneling efficiency. The variation in magnitude and the directionality of the temperature dependence of the primary KIE for the hydride in human DHFR are both consistent with the presence of an internal dynamic motion, or a promoting vibration as shown in Figure 5.³⁵ The observed break in the temperature dependence is indicative of a motion that can be “frozen out,” where below a certain temperature threshold the motion cannot contribute to the hydride transfer process. Above the thermal barrier, a compression of the internal axis provides an oscillatory modulation of the hydride donor–acceptor distance on the time scale of catalysis, either due to increased flexibility and thermal activation of the motion or an increase of its amplitude. The result of the motion being engaged is an increase in the efficiency of the compression between the hydride donor and acceptor that modifies the reaction coordinate.

Figure 5.

hsDHFR promoting vibration. Structure of hsDHFR (PDB 4M6J) with NADPH H-donor (blue) and folate acceptor (red). The PV residues determined by Schwartz and Masterson²⁰ are shown in purple, including I17, 24-PWPPLRNE-31, F32, and 62-PEKN-65. Black arrows indicate compressive motion of RPV.

Such a compressive motion has been proposed to exist in human DHFR (Figure 5),²⁰ and the direct measurement of temperature dependent KIE values are the first experimental evidence to support this class of dynamics in the enzyme. Computational analysis of DHFR from E. coli did not reveal a promoting vibration, supported by the consistent KIE across a range of temperatures.¹⁸ A promoting vibration coupled with the reaction coordinate should decrease the KIE as the compressive motion both decreases the donor–acceptor distance and compresses the barrier;³⁴ if the motion is not present, the KIE remains constant. Likewise, if the promoting vibration is unable to contribute due to a low temperature, as in hsDHFR at 5 °C, the KIE is equivalent to that of the ecDHFR.

Temperature dependent KIEs have been observed in other enzymes thought to contain promoting vibrations, with the same inverse proportionality with temperature as those observed in the present study.^23,24 Most notably, thermophilic alcohol dehydrogenase demonstrates more of a “break” in KIE, where the KIE is consistent below a certain temperature, and then becomes temperature dependent. It is clear from the data presented here that hsDHFR also exhibits a break rather than a continual dependence, with the “break” point falling between 15 and 25 °C. Yeast alcohol dehydrogenase specifically has been investigated and computationally found to also contain a donor–acceptor compressive motion.³⁵

Kohen and co-workers observed a break in KIE behavior for isotopically labeled heavy hsDHFR, absent for the light version. A steep increase in the intrinsic H/T KIE with decreasing temperature, while not derived from single-turnover experiments, was attributed by the authors to a semiclassical transition state theory model that incorporated a moderate contribution from tunneling or, due to observed calculated differences in the donor–acceptor distance, a Marcus-like model where this distance varied among different enzyme conformers.^16,23 It is not clear how this model accounts for the absence of the break in KIE in the light enzyme.

CONCLUSIONS

The kinetic study presented here exploits a sub-millisecond mixing time stopped-flow instrument to directly observe the pre-steady-state hydride transfer in two species of DHFR. For DHFR from E. coli, the maximum hydride transfer rate at low pH was updated to approximately 1100 s⁻¹, with direct measurements made at pH 5.0 that exceed the previous limit of 950 s⁻¹. The pK_a of the Asp27 side chain, the acidic residue that donates the proton for the reduction of dihydrofolate to tetrahydrofolate, was also estimated to be more acidic than previously established. Hydride transfer kinetics for human DHFR were thoroughly investigated for the first time using this improved dead time, and the rate at neutral pH was found to be approximately 10-fold greater than that for the E. coli enzyme. The established rate maximum of 4600 s⁻¹ is likely a low estimate, as acidification past pH 7 accelerated the hydride transfer rate to the limit of the capabilities of the instrumentation. Our findings are in good agreement with the simulated hydride transfer rates for both enzymes at 25 °C and pH 7, 220 ± 10 and 1100 ± 10 s⁻¹ for ec- and hsDHFR, respectively.³³

In addition to the basic kinetic characterization of pH dependence, the ability to measure the hydride transfer in human DHFR allowed the determination of the temperature-dependent primary kinetic isotope effects, providing experimental support for the presence of fast enzyme dynamics coupled to catalysis. The presence of a temperature dependent KIE in DHFR from H. sapiens (and absence in the E. coli counterpart) supports computational predictions that the more evolved active site contains a promoting vibration that modulates the hydride donor–acceptor distance.²⁰

While there exists no direct evidence for tunneling in any species of DHFR and tunneling-dependent processes do generally result in higher H/D KIE values than what is observed for DHFR, the kinetic scheme for this enzyme is complex and even with the ability to isolate a single turnover rate there is no certainty of isolating a single chemical event. It is also possible that the temperature dependent phenomena are associated with more global conformational changes not explicitly related to barrier crossing, such as the precatalytic search for reactive confirmations. Either interpretation highlights the critical role of protein dynamics in enzyme catalysis, and the possible presence of a promoting vibration in human DHFR also opens the door for allosteric inhibition by modulation of the protein dynamics.^36,37

ABBREVIATIONS

DHF

dihydrofolate

DHFR

dihydrofolate reductase

DNA

deoxyribonucleic acid

ec

Escherichia coli

FPLC

fast protein liquid chromatography

FRET

Förster resonance energy transfer

HPLC

high performance liquid chromatography

hs

Homo sapiens

IPTG

isopropyl β-d-thiogalactopyranoside

KIE

kinetic isotope effect

LB

Luria–Bertani

NADPH

reduced nicotinamide adenine dinucleotide phosphate

NMR

nuclear magnetic resonance

TEV

tobacco etch virus