The Reaction of the Substrate Analog 2-Ketoglutarate with Adenosylcobalamin-dependent Glutamate Mutase*

Glutamate mutase is one of several adenosylcobalamin-dependent enzymes that catalyze unusual rearrangements that proceed through a mechanism involving free radical intermediates. The enzyme exhibits remarkable specificity, and so far no molecules other than l-glutamate andl-threo-3-methylaspartate have been found to be substrates. Here we describe the reaction of glutamate mutase with the substrate analog, 2-ketoglutarate. Binding of 2-ketoglutarate (or its hydrate) to the holoenzyme elicits a change in the UV-visible spectrum consistent with the formation of cob(II)alamin on the enzyme. 2-ketoglutarate undergoes rapid exchange of tritium between the 5′-position of the coenzyme and C-4 of 2-ketoglutarate, consistent with the formation of a 2-ketoglutaryl radical analogous to that formed with glutamate. Under aerobic conditions this leads to the slow inactivation of the enzyme, presumably through reaction of free radical species with oxygen. Despite the formation of a substrate-like radical, no rearrangement of 2-ketoglutarate to 3-methyloxalacetate could be detected. The results indicate that formation of the C-4 radical of 2-ketoglutarate is a facile process but that it does not undergo further reactions, suggesting that this may be a useful substrate analog with which to investigate the mechanism of coenzyme homolysis.

Glutamate mutase catalyzes the reversible isomerization of L-glutamate and L-threo-3-methylaspartate (1)(2)(3). It is one of a group of enzymes that use adenosylcobalamin (AdoCbl), 1 a biologically active form of vitamin B 12 , to catalyze unusual 1,2-rearrangements in which an electron-withdrawing group is interchanged with a hydrogen atom on an adjacent carbon (4 -6) The migrating group may be -OH, -NH 2 , or, as in the case of glutamate mutase, a carbon-containing fragment so that a skeletal rearrangement is effected (Fig. 1). The role of AdoCbl as the source of carbon-based radicals, which are unmasked by homolysis of the coenzyme cobalt-carbon bond, is well established (7). Experiments with isotopically labeled substrates and coenzyme have demonstrated that for all the enzymes examined including glutamate mutase, 5Ј-deoxyadenosine acts as the intermediate carrier of the migrating hydrogen (8,9).
Recently, we have shown that in glutamate mutase homolysis of the cobalt-carbon bond and hydrogen abstraction from the substrate are kinetically coupled processes (10). Adenosyl rad-ical can therefore only be formed as a transient high energy species, or may not be formed at all if coenzyme homolysis and hydrogen abstraction are truly concerted processes. Similar findings have been reported for the related AdoCbl-dependent enzyme, methylmalonyl-CoA mutase (11). These unexpected observations have focused our attention on the role of the substrate in initiating free radical formation. A further intriguing and poorly understood aspect of these reactions is how the enzyme controls the rearrangement of substrate-radical to product-radical.
In principle, these steps might be probed with substrates that give rise to radicals of differing stabilities or that only participate in part of the catalytic cycle. However, attempts to find alternative substrates for glutamate mutase have been unsuccessful because the enzyme, in common with most other AdoCbl-dependent isomerases, appears to show very high substrate specificity. Several glutamate analogs have been reported to act as competitive inhibitors of the enzyme, among them (2S,3S)-and (2S,3R)-3-methylglutamate (12), (2R,3RS)-3-fluoroglutamate, (2S,4S)-4-fluoroglutamate (13), and 2methyleneglutarate (13) which, interestingly, is a substrate for the related AdoCbl-dependent enzyme, 2-methyleneglutarate mutase (14).
We report experiments with the glutamate analog 2-ketoglutarate (2-KG). We show that 2-KG functions in a partial reaction with the enzyme that results in the rapid transfer of tritium between the 5Ј-carbon of AdoCbl and C-4 of 2-KG. However, the resulting 2-ketoglutaryl radical appears unable to proceed onwards to the expected branch-chain product; instead, under aerobic conditions, a relatively slow side reaction occurs that results in inactivation of the enzyme.

EXPERIMENTAL PROCEDURES
Materials-The purification of glutamate mutase apoenzyme, GlmES, has been described previously (15). 3-Methylaspartase was purified as described by Hsiang and Bright (16). 5Ј-[ 3 H]AdoCbl was synthesized by using glutamate mutase to catalyze the exchange of tritium from tritiated glutamate to AdoCbl (9). AdoCbl and 2-KG were purchased from Sigma; tritiated glutamic acid was purchased from Amersham PLC. The sources of other materials have been described previously (9,15) or were purchased from commercial suppliers.
Inactivation of Glutamate Mutase by 2-KG-Reactions were set up in small vials containing 10 M GlmES, 10 M AdoCbl, and 10 mM 2-KG in 50 mM potassium phosphate buffer, pH 7.0 at 25°C in the dark. At various times, portions of the reaction mixture were withdrawn and divided in two; one portion was assayed for enzyme activity in the presence of fresh coenzyme; the other portion was used to determine the amount of AdoCbl remaining in the reaction by HPLC. For anaerobic work the vials were fitted with septa and made anaerobic by repeated evacuation and flushing with argon; transfer of reagents between vials was made by syringe.
UV-visible Spectra of Enzyme-Substrate Complexes-A solution containing 50 M glutamate mutase and 50 M AdoCbl in 50 mM potassium phosphate buffer, pH 7.0, was made anaerobic by repeated evacuation and flushing with argon. The solution was introduced by syringe into a cuvette fitted with a septum. The spectrum of the resting enzyme was recorded using a Hewlett-Packard diode array spectrometer. A concen-trated anaerobic solution of either 2-KG or L-glutamate was then added (final concentration 5 mM), and after 20 s, the spectrum of the enzymesubstrate complex recorded.
Tritium Exchange into 2-KG-Reactions were set up at 25°C under argon in light proofed 2.5-ml glass vials fitted with septa. Reaction mixtures (final volume, 0.5 ml) contained 10 M GlmES protein, 10 M 5Ј-[ 3 H]AdoCbl (100,000 dpm/nmol) in 50 mM potassium phosphate buffer, pH 7.0. The reaction was initiated by addition of 50 l of 100 mM 2-KG to give a final concentration of 10 mM. At various times, 50-l portions of the mixture were withdrawn, and the reaction was quenched by the addition of an equal volume of 5% trifluoroacetic acid solution. The reaction products were separated by HPLC on a C 4 reverse phase column as described previously (9), and the tritium content of the recovered 2-KG and AdoCbl was determined by scintillation counting.
Position of Tritium within 2-KG-A sample containing 10 M glutamate mutase, 10 M 5Ј-[ 3 H]AdoCbl (10,000 dpm total activity) and 10 mM 2-KG in 200 l of 50 mM potassium phosphate buffer, pH 7.0, was allowed to react for 45 min at room temperature, after which time exchange of tritium into 2-KG was essentially complete. AdoCbl was removed from the mixture by adsorption onto activated charcoal, and the sample was divided into two portions: one-half was brought to pH 10.2 by addition of sodium bicarbonate buffer, and as a control, the other half was maintained at pH 7.0. The samples were allowed to equilibrate for 24 h and then the water was removed by bulb-to-bulb distillation under vacuum. The tritium contents of the distilled water and the nonvolatile fractions were determined by scintillation counting.
NMR Spectroscopy-A 0.5 ml solution containing 50 M enzyme, 50 M AdoCbl, 50 mM potassium phosphate buffer, pH 7.0, in D 2 O was made anaerobic and introduced into an NMR tube. An anaerobic solution of 2-KG in D 2 O was added to a final concentration of 10 mM, and the sample was incubated at room temperature in the dark. At various times over the following 24 h, NMR spectra of the sample were acquired. Spectra were recorded at 400 MHz. Enzyme Assay-Glutamate mutase activity was measured spectroscopically by coupling the formation of 3-methylaspartate to mesaconate through the action of methylaspartase, as described by Barker et al. (1), except that the buffer was 100 mM potassium phosphate at pH 7.0.
Data Analysis-The data were plotted and curves fitted using the KaleidaGraph TM program (Abelbeck Software).

RESULTS
Our experiments used an engineered form of glutamate mutase, GlmES, in which the weakly associating E and S subunits of the wild-type enzyme have been genetically fused into one protein chain by introduction of an eleven amino acid linker sequence (15). This enzyme catalyzes the reaction with an efficiency similar to that of wild-type enzyme, but it binds coenzyme stoichiometrically and in a manner that is independent of protein concentration. These properties of the GlmES protein greatly simplify interpretation of mechanistic experiments (10,15).
Inactivation of Glutamate Mutase by 2-KG-Our initial experiments focused on whether 2-KG or the hydrated form that predominates in solution acted as either a competitive inhibitor or an irreversible inhibitor of glutamate mutase. 2-KG did not appear to bind very tightly to the enzyme, as no inhibition of activity was observed at concentrations up to 1 mM; for comparison, the K m for L-glutamate is 0.5 mM. It was not possible to extend measurements to higher concentrations of 2-KG as the compound absorbs at 240 nm, the wavelength used to assay for glutamate mutase activity.
Despite the apparently weak binding of 2-KG, upon incubation with glutamate mutase and AdoCbl over the course of several hours the compound was found to irreversibly inactivate the enzyme (Fig. 2). The initial loss of activity was well described by pseudo-first order kinetics with k inact ϭ 0.058 Ϯ 0.008 min Ϫ1 and was accompanied by degradation of the coenzyme, which occurred with essentially identical kinetics, k ϭ 0.061 Ϯ 0.008 min Ϫ1 , suggesting that the two processes were mechanistically linked. Control experiments established that neither AdoCbl nor 2-KG alone caused significant loss of enzyme activity, indicating that inactivation results from reaction of the coenzyme and substrate on the protein. However, very little loss of activity was observed when the experiments were repeated under rigorously anaerobic conditions (Fig. 2).
Taken together, these observations suggested that, upon binding to the holoenzyme, 2-KG was able to initiate homolysis of AdoCbl and that inactivation resulted from the subsequent reaction of oxygen with free radical species. It was apparent that neither all the coenzyme was degraded nor all the activity lost. This phenomenon may be attributed to the dilution of both active enzyme and AdoCbl that occurs during the reaction, which, in turn, results in dissociation of the enzyme⅐AdoCbl complex so that at low concentrations the reaction is no longer pseudo-first order.
UV-visible Spectrum of the Holoenzyme⅐2-KG Complex-When glutamate mutase holoenzyme is mixed with either Lglutamate or L-threo-3-methylaspartate, changes in the UVvisible spectrum of AdoCbl are observed that are associated with the formation of Co(II) species on the enzyme (10,17). We therefore examined whether 2-KG initiated similar spectral changes indicative of Cbl(II) formation, as would be expected if radical species accumulated on the enzyme. Addition of 2-KG (final concentration 5 mM) resulted in a decrease in absorbance at 520 nm and an increase in absorbance at 470 nm (Fig. 3 ), consistent with the formation of Cbl(II) species on the enzyme. Increasing the concentration of 2-KG to 25 mM did not result in any further significant changes in the spectrum. Assuming ⌬⑀ 530 of 4,000 M Ϫ1 cm Ϫ1 , (10) the proportion of enzyme in the Cbl(II) form is 35%, whereas in the presence of 5 mM L-glutamate, 45% of the enzyme is in the Cbl(II) form.
Tritium Exchange between AdoCbl and 2-KG-To seek further evidence that 2-KG was directly participating in the reaction mechanism of glutamate mutase, we investigated whether the enzyme could catalyze the transfer of tritium from the 5Ј position of AdoCbl to 2-KG. When the enzyme and 5Ј-[ 3 H]AdoCbl (10 M each) were incubated with 10 mM 2-KG, tritium was rapidly lost from AdoCbl and correspondingly ap-peared in 2-KG (Fig. 4). The time course of these two processes mirror each other, consistent with tritium being transferred directly from AdoCbl to 2-KG. The time course of the transfer reaction appeared to be biphasic; a rapid initial phase was observed that was complete within 20 s, followed by a slower phase that occurred over several minutes. Similar biphasic kinetic behavior was observed previously for the transfer of tritium between AdoCbl and the natural substrate, 3-methylaspartate (9). In that case it was shown that the fast phase of the reaction was due to the transfer of tritium from enzyme-bound AdoCbl to methylaspartate, whereas the slower phase represented exchange of bound coenzyme with unbound coenzyme. A similar phenomenon would explain the kinetic behavior observed in this case. The dissociation constant for AdoCbl is 2 M; therefore only about 75% of the coenzyme is expected to be bound to the enzyme under the conditions of the experiment. This is in accord with the relative amplitudes observed for the two kinetic phases. The apparent rate constant for tritium transfer from AdoCbl to 2-KG was 0.3 s Ϫ1 . This compares favorably with the value of 0.5 Ϯ 0.05 s Ϫ1 measured previously for tritium transfer from AdoCbl to methylaspartate (9).
Position of Tritium in 2-KG-To establish whether tritium was transferred to C-3 or C-4 of 2-KG, we exploited the fact that the C-3 protons readily exchange with solvent under alkaline conditions, and therefore any tritium at this position should be lost to water. The enzyme, 5Ј-tritiated AdoCbl, and a large excess of 2-KG were incubated together, and after tritium exchange was complete, AdoCbl was removed by adsorption on charcoal. The pH of the sample was then raised to 10.2, and the C-3 protons were allowed to exchange with solvent. However, after distillation no tritium was found in the water, and all the radioactivity was accounted for in the nonvolatile fraction. We conclude, therefore, that the tritium is exclusively introduced at C-4 of 2-KG, in accord with mechanistic expectations based on the reaction of the enzyme with glutamate. This observation also excludes the transient formation of 3-methyloxalacetate at the active site (followed by reversal to 2-KG), as this would result in scrambling of the tritium between C-3 and C-4 of 2-KG.
Attempts to Detect Turnover of 2-KG-The experiments described above all pointed to the C-4 radical of 2-KG being formed at the enzyme active site. It therefore seemed plausible that 2-KG might be converted into 3-methyloxalacetate in a rearrangement analogous to that undergone by the natural substrate. NMR spectroscopy was used to examine whether 2-KG was being converted into 3-methyloxalacetate or other compounds by the enzyme. No significant changes in the NMR spectrum of the sample were apparent after 24 h. In particular, the rearranged product should contain a methyl group that would give rise to a characteristic doublet in the high field region of the spectrum, but no resonances were evident in this region.
We conservatively estimate that we could detect about 2% conversion of substrate to product under the experimental conditions employed. This would correspond to only four turnovers of the enzyme in 24 h; for comparison, k cat for GlmES is 5 s Ϫ1 at 25°C (15). This result indicates that very little if any of the 2-ketoglutaryl radical generated on the enzyme undergoes rearrangement.

DISCUSSION
The AdoCbl-dependent isomerases in general show very high substrate specificity. This has hampered efforts to gain insight into the mechanism of these enigmatic rearrangements by using substrate analogs that might potentially function as mechanism-based inhibitors of these enzymes. Particularly intrigu- ing is the ability of these enzymes to stabilize free radical species, formed through homolysis of AdoCbl, by an estimated 25-30 kcal/mol relative to their formation in free solution (18,19). Our experiments demonstrate that 2-KG participates in a partial catalytic cycle on the enzyme that results in the formation of the C-4 radical of either 2-KG or its hydrate, however, this radical does not appear to undergo rearrangement to give products.
2-KG is the first molecule, other than the true substrates, for which the reversible transfer of hydrogen between coenzyme and substrate, a crucial step in the mechanism of the enzyme, has been demonstrated. The tritium exchange reaction is rapid, approaching the rate observed previously with methylaspartate (9), indicating that it serves as a proficient substrate for the hydrogen transfer reaction. We have not established the stereochemistry of hydrogen abstraction at C-4 of 2-KG, but it is expected that the pro-S hydrogen is exchanged with AdoCbl by analogy with the stereochemical course of the reaction with L-glutamate. Although incubation with 2-KG under aerobic conditions does lead to inactivation of the enzyme, this appears to be a slow side reaction that is also observed with the natural substrates (13). Oxygen, presumably, reacts with radical species at the enzyme active site, leading to oxidation of the coenzyme and/or the protein that results in inactivation.
It remains unclear whether it is the keto or hydrated form of 2-KG (or both) that is bound by the enzyme. Under the conditions of the reaction the hydrated form predominates, but there is no strong evidence to favor one species over the other, as either can be considered a reasonable mimic of glutamate and could in principle undergo hydrogen abstraction. However, hydration would make a difference in the ability of the substrate radical to undergo rearrangement. In the keto form, the substrate radical could, in principle, rearrange through the intermediacy of a cyclopropyl radical in which the unpaired electron is delocalized onto the oxygen: a precedent for such a rearrangement comes from the isomerization of succinyl-CoA analog, succinyl-carba-(dethia)-CoA by methylmalonyl-CoA mutase (20). In the hydrated form, the migrating carbon would be sp 3 hybridized and therefore unable to rearrangement through a cyclopropyl intermediate. Instead a fragmentation-recombination mechanism, analogous to that proposed for the rearrangement of glutamate to methylaspartate (5) would have to apply. This would involve the formation of an oxalyl radical intermediate, as opposed to a glycyl radical (as shown in Fig. 1), which may be less energetically favorable.
The UV-visible spectrum (Fig. 3) indicates that about 35% of the holoenzyme exists in the Cbl(II) form when 2-KG is bound, this is similar to the amount of coenzyme cleavage observed with glutamate bound (45%). The identity of the free radical partner to Cbl(II) formed in the reaction of 2-KG with glutamate mutase is of interest. Buckel and co-workers (21) have recently identified the C-4 radical of glutamate as the organic radical that accumulates on the enzyme during catalysis. Furthermore, we have demonstrated that, for both glutamate and methylaspartate, coenzyme homolysis and hydrogen abstraction are kinetically coupled (10). It seems most likely, therefore, that the products of homolysis in the present case are Cbl(II), 5Ј-deoxyadenosine and 2-ketoglutaryl radical (or its hydrate), as opposed to 5Ј-deoxyadenosyl radical and 2-KG.
The mechanism by which AdoCbl-dependent enzymes effect homolysis of the coenzyme, a reaction that is highly endothermic in free solution, to produce relatively stable radical species at the active site remains one of the most challenging aspects of B 12 catalysis. Pre-steady state kinetic analysis of the cobaltcarbon bond breaking step is complicated by the fact that the substrate radicals produced go on to rearrange so that multiple equilibria become established on the enzyme (10). In this regard, 2-KG may prove a very useful substrate for studying the coenzyme homolysis/hydrogen atom abstraction reaction because the 2-KG radical, once formed, does not react further. This should simplify the interpretation of kinetic experiments with isotopically labeled substrate and coenzyme that aim to gain insights into the transition state of this reaction.