Hydrogen tunneling in adenosylcobalamin-dependent glutamate mutase: evidence from intrinsic kinetic isotope effects measured by intra-molecular competition

Abstract

Hydrogen atom transfer reactions between substrate and coenzyme are a key mechanistic feature of all AdoCbl-dependent enzymes. For one of these enzymes, glutamate mutase, we have investigated whether hydrogen tunneling makes a significant contribution to the mechanism by examining the temperature-dependence of the deuterium kinetic isotope effect associated with hydrogen atom transfer from methylaspartate to the coenzyme. To do this we designed a novel intra-molecular competition experiment that allowed us to measure the intrinsic kinetic isotope effect, even though hydrogen transfer may not be rate determining. From the Arrhenius plot of the kinetic isotope effect, the ratio of the pre-exponential factors A_H/A_D was 0.17 ± 0.04 and the isotope effect on the activation energy, ΔE_{a(D – H)} was 1.94 ± 0.13 kcal/mol. The results imply that significant degree of hydrogen tunneling occurs in glutamate mutase, even though the intrinsic kinetic isotope effects are well within the semi-classical limit and are much smaller than those measured for other AdoCbl enzymes and model reactions for which hydrogen tunneling has been implicated.

Evidence supporting quantum mechanical tunneling of hydrogen has been obtained for an increasing number of enzyme reactions (1–9). The most common, and readily interpretable, test for hydrogen tunneling involves measuring the temperature-dependence of kinetic isotope effects (9, 10) (KIEs)¹, although other kinetic isotope-based tests have also been used to diagnose hydrogen tunneling (11–13). Such experiments have subjected enzyme-catalyzed reactions to detailed scrutiny and required the development of improved theoretical descriptions to accurately describe the data (1, 2). This, in turn, has resulted in some of the most complete descriptions of enzymatic transition states and led to an increased understanding of the important role that protein dynamics plays in catalysis.

In this report we present data to substantiate hydrogen tunneling in the AdoCbl-dependent enzyme glutamate mutase, which catalyzes the radical-mediated rearrangement of L-glutamate to L-threo-3-methylaspartate (14–18). Hydrogen atom transfer between the substrate and coenzyme is a key step in the mechanisms of all AdoCbl-dependent enzymes (19, 20): first, homolysis of the AdoCbl cobalt-carbon bond transiently generates 5′-deoxyadenosyl radical; next, this highly unstable radical immediately reacts to abstract a hydrogen atom from the substrate, thereby generating a substrate radical that undergoes subsequent steps in the mechanism.

Evidence to support hydrogen tunneling in AdoCbl reactions has been obtained from the temperature dependence of KIEs measured in AdoCbl-dependent methylmalonyl-CoA mutase (21) and from model studies on the thermolysis of AdoCbl and neopentyl-Cbl in protiated and deuterated ethylene glycol (22, 23). In these studies very large deuterium KIEs (ranging from ~ 20 to ~ 50) were reported that exceed the semi-classical limit of ~ 7 expected in the absence of tunneling. However, we recently measured the intrinsic KIE for the transfer of hydrogen between methylaspartate and 5′-dA in glutamate mutase (24) and found it to be much smaller, KIE ~ 4 at 10 °C, which is clearly well within the semi-classical limit. Thus it was unclear whether hydrogen transfer in glutamate mutase also involves a significant degree of quantum tunneling, as appears to be the case in methylmalonyl-CoA mutase.

In this paper we have determined the temperature dependence of the intrinsic deuterium KIE for the transfer of hydrogen from methylaspartate to 5′-dA, catalyzed by glutamate mutase. The KIE was found to be temperature dependent, leading to an apparent isotope effect on the Arrhenius pre-exponential factors, A_H/A_D, of 0.17 ± 0.04 and an isotope effect on the activation energy, ΔE_{a(D – H)}, of 1.94 ± 0.13 kcal/mol. These parameters are different from those measured previously for methylmalonyl-CoA mutase and the thermolysis of model alkyl-cobalamins (21–23); in both of these cases A_H/A_D was smaller and ΔE_{a(D – H)} was larger.

Experimental Procedures

Reagents

Synthesis of d₁-methyl-(2S,3S)-methylaspartate and purification of glutamate mutase from recombinant E. coli have been described previously.(25, 26) All other reagents were of the highest quality commercially available. We note that these experiments were all performed with the GlmES fusion protein in which the weakly associating S and E subunits of glutamate mutase have been genetically fused together; this simplifies kinetic analyses by removing protein concentration dependence from the kinetics of the reaction (25).

Rapid quench experiments were performed at various temperatures between − 2.5 °C (the lowest temperature accessible without needing to include cryoprotectants to prevent freezing) and 30 °C using a Hi-Tech RQF-63 apparatus. 73.5 μL of 275 μM glutamate mutase in 50 mM potassium phosphate buffer, pH 7.0 was immediately mixed with 6.5 μL 3.4 mM AdoCbl immediately before the experiment to reconstitute the holoenzyme. The holoenzyme mixture was reacted with 80 μL of 2.2 mM d₁-methylaspartate in 50 mM potassium phosphate buffer, pH 7.0. The reactions were quenched with 80 μL of 2.5 % trifluoroacetic acid, after aging for various times (12 – 160 ms). 5′-dA was recovered and quantified by C₁₈ reverse-phase HPLC as described previously (27). All the procedures were conducted in dim red light. Samples of labeled 5′-dA were lyophilized and stored at −20 °C prior to mass spectrometric analysis.

Mass spectral analysis

5′-dA was re-dissolved in solvent containing 50% MeOH and 0.1% formic acid. Experiments were performed in positive ion mode at 50 μL/h with a 7 T Q-FT-ICR mass spectrometer (Bruker Daltonics, Billerica, MA) as described previously (28). Data were collected in narrowband mode. The masses of ¹²C₉ ¹H₁₃ ¹³C₁ ¹⁴N₅ ¹⁶O₃ 5′-dA and ¹²C₁₀ ¹H₁₂ ²H₁ ¹⁴N₅ ¹⁶O₃ 5′-dA are 252.10436Da and 252.10799Da, respectively: a mass difference of 3.6 mDa. The experimental masses of the enzymatic samples compare accurately (<2ppm) with the masses of the standards and the calculated theoretical masses. As an internal control, the instrument was tuned to ensure that the detected relative abundance of ¹²C-5′-dA to ¹³C-5′-dA was as close as possible to the calculated ratio of 0.1325.

Calculation of intrinsic KIE

The intrinsic KIE was calculated either by integrating the peak areas for ¹²C-¹H-5′-dA and ¹²C,²H-5′-dA and directly comparing their ratio. A statistical factor of 2 was used to account for the fact that the labeled methylaspartate contains two protium atoms and only one deuterium atom in the abstractable position. Alternatively, because the natural abundance of ¹³C is known to great precision, the intrinsic KIE was also calculated by comparing the ratio of ¹³C-¹H-5′-dA to ¹²C,-²H-5′dA, and correcting the result for the natural abundance, n, using the equation below:

$$\mathit{KIE}=\frac{1}{2}\times \frac{\left[{}^{13}C-5^{\prime }dA\right]}{\left[{}^{2}H-5^{\prime }dA\right]}\times \frac{(1-n)}{n}$$

The KIEs calculated by either method were the same within experimental error.

Results and Discussion

To reliably diagnose hydrogen tunnelling in enzymes, it is imperative to measure the intrinsic KIE for the hydrogen transfer step, undiminished by other steps in the mechanism that may be slower. Whereas for some enzymes there is a single isotopically sensitive step that is cleanly rate-determining in the mechanism, the mechanisms of AdoCbl-dependent rearrangements require two hydrogen transfer steps. Previous experiments on glutamate mutase that have employed a combination of kinetic measurements to elucidate the free energy profile of the reaction indicate that neither hydrogen transfer step is cleanly rate-determining (18, 27, 29–34). Pre-steady state techniques often allow the intrinsic KIE to be measured, but if a slow step precedes the isotopically sensitive step, the observed KIE will still be diminished from its intrinsic value. Results from various pre-steady state measurements using UV-visible stopped-flow spectroscopy and rapid chemical quench techniques suggest that this situation may well apply in glutamate mutase (18, 27, 29, 32, 33).

To overcome this problem, we developed a method to measure the intrinsic KIE in glutamate mutase that relies on intra-molecular competition (35) between hydrogen and deuterium atoms present in the same methyl group of the substrate, methylaspartate (24). As illustrated in Figure 1, during the reaction 5′-deoxyadenosyl radical, generated by homolysis of AdoCbl, is confronted with the choice of abstracting either protium or deuterium from the methyl group of the same substrate molecule. The isotopic composition of 5′-dA formed in the reaction therefore reflects the intrinsic KIE. The intrinsic KIE² can be measured, even when the isotopically sensitive step is not rate determining, because it is manifested though intra-molecular competition between protium and deuterium atoms, which remain chemically equivalent, even in the enzyme active site, due to the rapid rotation of the methyl group.

Figure 1.

Measurement of intrinsic KIE for hydrogen atom transfer from methylaspartate to 5′-dA in glutamate mutase by intra-molecular competition between deuterium and hydrogen atoms at the methyl group of the substrate.

The measurement of KIEs by intra-molecular competition followed the general method that we described previously (24). Glutamate mutase and [d₁-methyl]-(2S,3S)-3-methylaspartate were reacted together at various temperatures using a rapid quench flow apparatus, as described in the experimental procedures section. The solutions were mixed in a 1:1 ratio and allowed to age for various times (10 – 120 ms) before being quenched and the 5′-dA formed in the reaction was recovered by reverse-phase HPLC (27). The deuterium content of the 5′-dA was then determined by ultra high-resolution Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR-MS). The resolution of this type of mass spectrometer is sufficiently high (36, 37) that the mixture of 5′-dA isotopomers, due to natural abundance ¹³C-, and ¹⁵N-containing molecules, as well as ²H- from the substrate, are baseline resolved as shown in Figure 2. Normally, the presence of ¹³C complicates the analysis of deuterium KIEs because the peaks due to deuterated and ¹³C-containing molecules overlap; in this case ultra-high resolution allows the ¹³C-peak to be used as an internal standard to measure the relative abundance of the peak due to ²H-labelled 5′-dA.

Figure 2.

Representative FT-ICR-MS spectrum of a sample of 5′-dA obtained after reacting glutamate mutase with [d₁-methyl]-methylaspartate for 44 ms at 2.5 °C. The region shown illustrates the separation of ¹⁵N-, ¹³C-, and ²H-containing isotopomers of 5′-dA.

The mechanism of the glutamate mutase-catalyzed reaction involves two hydrogen transfer reactions, one from substrate to coenzyme and the other from coenzyme to product. Therefore measurements have to be made under pre-steady state conditions; otherwise multiple turn-overs would lead to accumulation of deuterium in 5′-dA and a diminution of the observed KIE. To insure that this condition was met, we measured the KIE for a range of reaction times at each temperature we studied. As shown in Figure 3, the apparent KIE diminished as the reaction time was lengthened, but at times shorter than ~ 25 ms it was constant within experimental error. This result is consistent with previous measurements of the rate of 5′-dA formation that show a t_1/2 of ~ 11 ms at 10 °C (27). Therefore, within this short time span the measured KIE may be assumed to be close to the intrinsic KIE (38).

Figure 3.

Apparent decrease in KIE as a function of reaction time due to multiple turn-overs of the enzyme and equilibration of isotopes. The KIE plateaus for reaction times shorter than 25 ms, indicating that the intrinsic KIE is being measured at short reaction times. Data shown is for reaction at 5 °C.

The KIE was measured at various temperatures between − 2.5 °C and 30 °C. However, at temperatures above 10 °C increasing amounts of di-deuterated 5′-dA were detected, even at short reaction times, indicating that the reaction becomes too fast for the intrinsic KIE to be reliably measured (Figure 4). There is a low background amount of di-deuterated 5′-dA even at lower temperatures, which is expected due to natural abundance deuterium in the sample, but which is hard to quantify accurately.

Figure 4.

Increase in the formation of di-deuterated 5′-dA at higher temperatures, indicative of multiple turn-overs. Each data point represents the average of at least 3 measurements made with reaction times between 12 and 24 ms.

The Arrhenius plot for KIEs measured between − 2.5 °C and 10.5 °C is shown in Figure 5. Each point represents an average of at least five independent measurements of the KIE made at reaction times between 12 and 24 ms. It is evident that at higher temperatures the KIE departs significantly from the Arrhenius equation. This may be attributed to contribution from multiple transits of hydrogen between substrate and coenzyme as discussed above. However, between −2.5 °C and 5.0 °C the plot is linear, and from these data the apparent isotope effect on the pre-exponential factor, A_H/A_D, was 0.17 ± 0.04 and the isotope effect on the energy of activation, ΔE_a was 1.94 ± 0.13 kcal/mol. We note that whereas we were only able to study the KIE over a relatively narrow temperature range, it is not necessary to make measurements over a wide temperature range to reliably diagnose hydrogen tunneling. This is because the method amounts to drawing a tangent to the full temperature-dependent curve (10), rather than trying to extrapolate the data to infinite temperature to obtain the “true” pre-exponential factor.

Figure 5.

Arrhenius plot of the KIE for the formation of 5′-dA from methylaspartate and AdoCbl catalyzed by glutamate mutase over the temperature range −2.5 to 10.0 °C. Each data point represents the average of at least 5 measurements made with reaction times between 12 and 24 ms. The deviation from linearity at higher temperatures may be due to multiple transits of hydrogen at higher temperatures which reduce the apparent KIE.

The simplest interpretation of these data, in which the KIE is highly temperature dependent, is that the enzyme exhibits “moderate” tunneling; i.e. that protium is tunneling in the reaction whereas there is reduced tunneling for deuterium (in contrast, temperature independent KIEs indicate extensive tunneling of both protium and deuterium). More sophisticated analyses, that incorporate “Marcus-like” or “full tunneling” models to describe the temperature dependence of KIEs in enzymes, emphasize the role of protein dynamics in modulating hydrogen transfer (1, 9, 39–41). The physical explanation for temperature-dependent isotope effects that emerges from the models is that the substrate and coenzyme are not optimally oriented in the active site for tunneling to occur, and that a thermal “gating” motion of the protein is needed to bring the reactants into optimal alignment for tunneling.

Hydrogen tunneling has been investigated for methylmalonyl-CoA mutase (21), an AdoCbl-dependent enzyme that catalyzes a carbon skeleton rearrangement very similar to glutamate mutase. In this case, KIEs were measured using stopped flow spectroscopy to follow the rate of AdoCbl homolysis when the enzyme was reacted with unlabeled and deuterated methylmalonyl-CoA. Tunneling has also been investigated for the abstraction of hydrogen from ethylene glycol by adenosyl radical generated by thermolysis of AdoCbl, which mimics the reaction catalyzed by AdoCbl-dependent dioldehydratase (22). A similar experiment was performed using neopentylCbl (23) which is sufficiently reactive that neopentyl radical can be generated by thermolysis at room temperature. For comparison, the relevant parameters measured for these systems are collected together in Table 1. It is evident that the KIEs measured for these related reactions are an order of magnitude larger than those we determine for glutamate mutase. ΔE_{a(D – H)} is somewhat smaller for glutamate mutase, whereas A_H/A_D, which is determined least precisely in these experiments, is probably not significantly different between the different systems. Although the hydrogen atom donor is different in this case, methylaspartate as opposed to methylmalonyl-CoA or ethylene glycol, it would seem unlikely that this difference would, of itself, be responsible for substantially changing the intrinsic KIE. We also note that different experimental methods were used to measure the KIEs for each system, each of which may be subject to different types of systematic errors, but again we consider that this is an unlikely explanation for the differences in the observed KIEs.

Table 1.

Tunneling-related parameters determined for AdoCbl-dependent enzymes and model systems

Enzyme/model system	KH/KD (0 °C)	AH/AD	ΔEa(D – H) (kcal/mol)	Reference
Glutamate mutase	6.0	0.17 ± 0.04	1.94 ± 0.13	this work
Methylmalonyl-CoA mutase	~56a	0.08 ± 0.01	3.41 ± 0.07	(21)
AdoCbl + ethylene glycol	~40a	0.16 ± 0.07	3.0 ± 0.3	(22)
NeopentylCbl + ethylene glycol	~42a	0.14 ± 0.07	3.1 ± 0.3	(23)

^a KIEs extrapolated to 0 °C to facilitate comparison

Based on the similarity between the parameters listed in Table 1 for methylmalonyl-CoA mutase and the non-enzymatic systems it was argued that there was no additional contribution from the enzyme to tunneling; i.e. that the degree of hydrogen tunneling was the same in the enzyme catalyzed and uncatalyzed reactions (21–23). Subsequently, a computational study of methylmalonyl-CoA mutase was able to reproduce the unusually large experimental KIE through the use of a multidimensional tunneling model that included implicit protein motions in the reaction coordinate (42). To our knowledge, similar computational studies have not been performed to examine the KIEs measured for the AdoCbl model reactions.

Our results certainly point to an influence of the protein on the transition state for hydrogen atom transfer in glutamate mutase because the intrinsic KIE is much smaller than in either methylmalonyl-CoA mutase or the model systems. Full tunneling models predict that smaller KIEs are indicative of a better pre-organized active site in which the donor-acceptor distance at the tunneling-ready state is shorter.

Although in glutamate mutase the KIE is well within the limits expected for semi-classical behavior, the temperature dependence of the KIE points to hydrogen tunneling. This observation is in accord with our previous measurements of the secondary tritium KIEs associated with hydrogen transfer from glutamate to 5′dA (43, 44). In these experiments the enzyme was reacted with AdoCbl labeled with tritium at the 5′-carbon and either unlabelled or deuterated glutamate. It was found that the isotope at the primary position strongly influenced the magnitude of the secondary KIE, resulting in a breakdown in the rule of the geometric mean: i.e. that there are no isotope effects on isotope effects. It has been shown that this behavior is indicative of a transition state in which tunneling of hydrogen at the primary position is strongly coupled to the motion of the secondary hydrogen atoms (45, 46), in this case at the 5′-carbon of the adenosyl radical. Furthermore, this behavior has the effect of reducing the primary kinetic isotope effect, which is “stolen” by the coupled motion of the secondary hydrogens. Similar behavior has been observed in, for example, yeast alcohol dehydrogenase (47, 48) for which the primary KIEs are also well within the semi-classical range.

In conclusion, these results provide evidence for hydrogen tunneling in glutamate mutase. Together with the x-ray structure for the enzyme, they provide a starting point for computational modeling of the glutamate mutase reaction using various tunneling models. Such studies promise to provide detailed insights into the still poorly understood principles by which Ado-Cbl-dependent enzymes generate and control highly reactive free radical intermediates.