LETTERS

An allylic ketyl radical intermediate in clostridialamino-acid fermentationJihoe Kim1, Daniel J. Darley1{, Wolfgang Buckel1 & Antonio J. Pierik1

The human pathogenic bacterium Clostridium difficile thrives bythe fermentation of L-leucine to ammonia, CO2, 3-methylbutanoateand 4-methylpentanoate under anaerobic conditions1. The reduc-tive branch to 4-methylpentanoate proceeds by means of the dehyd-ration of (R)-2-hydroxy-4-methylpentanoyl-CoA to 4-methylpent-2-enoyl-CoA, which is chemically the most demanding step. Ketylradicals have been proposed2 to mediate this reaction catalysed byan iron–sulphur-cluster-containing dehydratase, which requiresactivation by ATP-dependent electron transfer from a secondiron–sulphur protein functionally similar to the iron protein ofnitrogenase. Here we identify a kinetically competent product-related allylic ketyl radical bound to the enzyme by electronparamagnetic resonance spectroscopy employing isotope-labelled(R)-2-hydroxy-4-methylpentanoyl-CoA species. We also found thatthe enzyme generated the stabilized pentadienoyl ketyl radical fromthe substrate analogue 2-hydroxypent-4-enoyl-CoA, supportingthe proposed mechanism. Our results imply that also other2-hydroxyacyl-CoA dehydratases3 and the related benzoyl-CoAreductases4—present in anaerobically living bacteria—employketyl radical intermediates. The absence of radical generators suchas coenzyme B12, S-adenosylmethionine or oxygen makes theseenzymes unprecedented in biochemistry.

Radical anions are versatile intermediates in enzymatic reactionsbecause they combine reactivity ‘Umpolung’5, which is polarityinversion, with acid–base catalysis. Thus, the transfer of one electroncan reduce an electrophilic carbonyl to a nucleophilic ketyl radical,which is stabilized by the negative charge of the oxygen interactingwith the half-filled 2pz orbital at the carbon6. Ketyl radicals eliminateadjacent leaving groups or are protonated to form neutral radicals. Inorganic chemistry, the reduction of a-hydroxyketones to unsubsti-tuted ketones by inorganic one-electron donors (Zn0, Cr21 or Sm21)initially converts the ketone to a ketyl capable of expelling the adja-cent hydroxyl group. The resulting enoxy radical is reduced to theenolate by a second electron and protonated to the ketone2. Similarly,the mechanism of several radical enzymes can be formulated to pro-ceed through ketyl radical intermediates, but experimental evidencefor this is lacking2,7.

Ketyl radicals have also been proposed as intermediates for2-hydroxyacyl-CoA dehydratases, which are a class of enzymesinvolved in the fermentation of 12 of the 20 proteinogenic aminoacids used by certain anaerobic bacteria1,3. The required eliminationof the b-hydrogen as a proton (pK < 40) and removal of the hydroxylgroup adjacent to a thioester carbonyl is, in terms of chemistry, a verychallenging reaction. It was proposed that the partial positive chargeof the carbonyl is reversed by one-electron reduction of the substrate(1; Fig. 1) to a substrate-derived ketyl radical (2), which expels thea-hydroxyl group. The resulting enoxy radical (3), with a much moreacidic b-hydrogen (pK < 14)8, is deprotonated by a base within the

enzyme, yielding the resonance-stabilized, product-related allylicketyl radical (4). A final oxidation leads to the product 2-enoyl-CoA (5) with the electron returning to the dehydratase for the nextturnover3.

The anaerobic human pathogenic bacterium Clostridium difficilecauses hospital-acquired diarrhoea, pseudomembranous colitis andoccasionally death9 by overgrowth in the gastrointestinal tract.Pathogenicity involves the synthesis of virulent toxins10, the produc-tion of noxious p-cresol from tyrosine11 and a large repertoire ofamino-acid fermentations12. Leucine, which is indispensable forgrowth, serves as both oxidant and reductant (Stickland reaction13);1 mol is oxidized to ammonia, 3-methylbutanoate and CO2, and 2 molare reduced to 4-methylpentanoate and ammonia14. In the reductivebranch, leucine is transaminated to 2-oxo-4-methylpentanoate,which is reduced by NADH and esterified by CoA transfer to (R)-2-hydroxy-4-methylpentanoyl-CoA15. The subsequent difficult dehy-dration is catalysed by two oxygen-sensitive proteins: a homodimericactivator and a heterodimeric dehydratase. Activation occurs byATP-hydrolysis-driven electron transfer from the readily reducible[4Fe–4S]1/[4Fe–4S]21 cluster of the activator to the [Fe–S] clusterof the dehydratase. After such a single activation step, the dehydratasecan recycle the electron for about 104 turnovers16. After the dehyd-ration to 4-methylpent-2-enoyl-CoA reduction to 4-methylpentanoyl-CoA and CoA transfer lead to liberation of the final product,4-methylpentanoate16.

1Laboratorium fur Mikrobiologie, Fachbereich Biologie, Philipps-Universitat, D35032 Marburg, Germany. {Present address: Department of Pharmacy and Pharmacology, University ofBath, Claverton Down, Bath BA2 7AY, UK.

O O

O O–O–

O

H

H

H H

HH

H

H

H

H

H

H

H+

CoAS

CoAS

CoAS

CoAS

CoAS

CoAS

Dehydratase

Activator

1

2

3

4

5H

H

H

H+

HO

HO

H2O

e–

e–

–

Figure 1 | Proposed mechanism for the enzymatic dehydration of (R)-2-hydroxy-4-methylpentanoyl-CoA. Substrate (1) is converted to 4-methylpent-2-enoyl-CoA (5) by 2-hydroxy-4-methylpentanoyl-CoA dehydratase and itsactivator from Clostridium difficile through a substrate-derived ketyl radical(2), an enoxy radical (3) and a product-related allylic ketyl radical (4), thesubject of this article, shown in two mesomeric forms. The activator is reducedby ferredoxin or flavodoxin in vivo, and by sodium dithionite in vitro.

Vol 452 | 13 March 2008 | doi:10.1038/nature06637

239Nature Publishing Group©2008

In the presence of activator, ADP and dithionite as in vitro reduc-tant, 2-hydroxy-4-methylpentanoyl-CoA dehydratase did not show asignal by electron paramagnetic resonance (EPR) spectroscopy in theS 5 1/2 region (g < 2; Fig. 2a, spectrum 1). The sample exhibited onlythe S 5 3/2 signals at g 5 4–6 (Fig. 2b, spectrum 1) of the [4Fe–4S]1

cluster of the activator17. Replacement of ADP by ATP reductivelyactivated the dehydratase, which showed a broad S 5 1/2 EPR signal(Fig. 2a, spectrum 2) with minor changes of the activator signals(Fig. 2b, spectrum 2). Because of the excess of dithionite, the activatorremained reduced. The g values (2.03 and 1.92), line width andbroadening beyond detection above 20 K are all characteristic fea-tures of other [4Fe–4S]1 cluster-containing proteins. On additionof the substrate (R)-2-hydroxy-4-methylpentanoyl-CoA to the acti-vated dehydratase, the line shape and g values in the EPR spectrum ofits [4Fe–4S]1 cluster changed substantially. We attribute the changein the [4Fe–4S]1 signal to the binding of substrate 1 or product 5 inthe non-radical form (Fig. 1). In parallel, a sharp EPR signal with anaverage g value, gav, of 2.0038 6 0.0002 (mean 6 s.e.m.) was formed(Fig. 2a, spectrum 3), which is reminiscent of signals observedwith organic radicals18. This signal has a microwave power of half-saturation at 10 K of 0.12 6 0.03 mW and broadened beyond detec-tion at temperatures above 60 K. Magnetic coupling of the radicalwith a paramagnetic state of the dehydratase (that is, [4Fe–4S]1) canbe excluded by the minor effect on relaxation and the narrowline width. In contrast, the paramagnetic excited state(s) of the[4Fe–4S]21 cluster at 5–8 A distance from organic radicals in lysine2,3-aminomutase19,20 and coproporphyrinogen III oxidase21 causeenhanced dipolar relaxation similar to that observed here.

Close inspection of the radical species in the g 5 2 region revealedtwo dominating hyperfine splittings from distinct I 5 1/2 nuclei withcoupling constants of 1.45 and 1.15 mT (Fig. 2c, non-labelled), whichare values typical for protons18. To determine the origin of thesetwo protons, a series of 2H-labelled substrates were prepared(Supplementary Methods, Supplementary Fig. 1 and Supplemen-tary Table 1) and the effect of labelling on the hyperfine couplingsof the EPR signal was determined (Fig. 2c). On 2H-labelling of theC-2 position (for numbering see Fig. 2d), the magnitude of the twohyperfine couplings remained unchanged and, with the exception ofa minor change of the line shape on the high-field side, the overall line

width of the EPR signal was also comparable. In contrast, 2H at theC-3 or C-4 position led to pronounced changes. In the case of 2H atC-3, the proton hyperfine coupling of 1.15 mT disappeared and asingle hyperfine coupling of 1.45 mT remained. On 2H-labelling ofthe C-4 position, exactly the opposite occurred: the 1.45-mT protonhyperfine coupling was lost and the second splitting was unaffected.Labelling at both the C-3 and C-4 positions abolished both the1.45-mT and 1.15-mT proton hyperfine couplings. Numeric simu-lation of the EPR spectra was performed with appropriate downscal-ing of proton hyperfine couplings by a factor of 6.514 (the 1H/2Hgyromagnetic ratio18). With the same parameters used for the non-labelled substrate, simulations of the EPR spectra of radical speciesobtained from substrates deuterated at C-3 and C-4 as well as at bothof these positions matched the experimental spectra (Fig. 2c, redtraces). Washout of 2H from both the C-3 and C-4 positions into aprotein or organic cofactor-derived radical species could in principleexplain our observations. Compelling evidence against such a‘secondary’ labelling situation was provided by the marked effect of(R)-2-hydroxy-4-methyl[1-13C]pentanoyl-CoA, obtained from 2-oxo-4-methyl[1-13C]pentanoate, on the EPR signal (Fig. 2c,[1-13C]). Successful simulation of the asymmetric EPR signalrequired a slightly rhombic g-tensor (gxyz 5 2.0049, 2.0038, 2.0027)and substantial anisotropy of the 13C hyperfine tensor (Axy 5 0.4 mT,Az 5 3.7 mT, Aiso 5 1.5 mT).

The observed hyperfine couplings define a delocalized spin densityminimally residing on carbon atoms 1 and 3. This notion eliminatesboth the substrate-derived ketyl radical (2, Fig. 1) and the enoxyradical (3, Fig. 1) as potential candidates, because neither of theseradicals can give rise to the experimentally observed 1.45-mT coup-ling with the proton at C-4. In contrast, the experimentally observed1H and 13C hyperfine couplings can be accounted for by the spindensities of the delocalized allylic ketyl radical species (4, Figs 1 and2d). In agreement with this assignment, the experimental gav

(2.0038 6 0.0002) is higher than those of all-carbon (hydroxy)allylicradicals (2.0025–2.0034) and closer to those of allylic O/S-substitutedketyl radicals in non-biological systems (2.0032–2.0044; Supplemen-tary Table 2). Hyperfine coupling constants of the proton attached tothe nodal central carbon atom range from 0.38 to 0.44 mT for all-carbon, or 0.05 to 0.36 mT for O/S-substituted planar allylic radicals

(1)

(2)

(3)

(1)

(2)

(3)

Non-labelled

300 330 360 390

50 100 150 200 250 300 331 334 337 340 343Magnetic field (mT) Magnetic field (mT)

[2-2H]

[3-2H2]

[4-2H]

[3,4-2H3]

[1-13C]

CoAS

H

H H

H

HH

H

(2)(1)

C-2

CH3

CH3

CH3

CH3

H3CH3C

C-2

O–

12 3 4

0 50 100 150 Time (ms)

2

1

Rad

ical

/deh

ydra

tase

(%)

a

b

c d

e

f

Figure 2 | EPR spectroscopy of the product-related allylic ketyl radicalintermediate in 2-hydroxy-4-methylpentanoyl-CoA dehydratase. a, b, Low-spin (a) and high-spin (b) EPR signals of [Fe–S] clusters and organic radical:2-hydroxy-4-methylpentanoyl-CoA dehydratase and its activator weremixed in the presence of ADP (spectra 1), ATP (spectra 2) or ATP and (R)-2-hydroxy-4-methylpentanoyl-CoA (spectra 3). c, EPR signals (black) of theorganic radical in activated dehydratase in the presence of ATP and (R)-2-hydroxy-4-methylpentanoyl-CoA isotopically labelled in the indicatedpositions (numbering as in d). Red traces are simulations of the

experimental spectra. d, Derived structure of the product-related allylic ketylradical. e, Newman projections along the C-3–C-4 bond of possibleconformers 1 and 2. f, Formation of the product-related allylic ketyl radical,as determined by rapid freeze-quench EPR spectroscopy. The solid line is anonlinear least-squared fit of the signal intensity (black circles) as a functionof time to a first-order rate equation (for spectra see Supplementary Fig. 3).For synthesis of substrates, sample preparation, EPR conditions andsimulation see Methods, Supplementary Text and Supplementary Fig. 1.

LETTERS NATURE | Vol 452 | 13 March 2008

240Nature Publishing Group©2008

(Supplementary Table 2). Thus, the observed lack of a resolved coup-ling of the proton at C-2 in our system is in perfect agreementwith the existence of an allylic ketyl radical. Support is also foundin the magnitude of the 13C hyperfine coupling constant and absenceof coupling with the CH2 moiety of CoA (see SupplementaryDiscussion and Supplementary Tables 3 and 4).

The conformation of the isopropyl moiety in the product-relatedradical bound to the dehydratase could be determined from theangular dependence of the hyperfine coupling of the 1H attachedto C-4. From the McConnell equation, A1

H 5 r(0.092 1 4.2 cos2h)(where h is the dihedral angle and r is the spin density, assumed to be0.5 6 0.1)19, and the 1.45-mT proton hyperfine coupling constant,four possible dihedral angles between the p orbital at C-3 and C4-Hcan be calculated: 135, –35, 1145 and –145u (all 610u). If we restricth to values for which energetically unfavourable steric repulsion ofthe methyl groups with C2-H of the allyl group is avoided, two con-formers are possible: 135 and 1145u (Fig. 2e, conformers 1 and 2,respectively).

For samples prepared under a broad variety of experimental con-ditions, a maximal allylic ketyl radical content of (2.1 6 0.4)% wascalculated from double integration and calibration with a Cu21 EPRstandard. At time scales from 20 to 250 s, the intermediate slowlydecayed, probably because of side reactions of the radical (Supple-mentary Fig. 2, about 0.004 s21). To assess the mechanistic relevanceof this observation, the rate of formation of this partly populatedintermediate was measured after mixing the activated dehydratasewith its natural substrate (R)-2-hydroxy-4-methylpentanoyl-CoA(Fig. 2f). The intermediate was trapped by rapid freeze-quenchingof the reaction mixture in liquid-nitrogen-cooled isopentane (about130 K). Even at the shortest reaction time point accessible (7 ms), thecharacteristic EPR signal of the allylic ketyl radical with spectralproperties indistinguishable from manually mixed samples (at least15 s) was detected (Supplementary Fig. 3). The same fractional popu-lation of the allylic ketyl radical was seen for all time points longerthan 40 ms. From the observed time dependence for the developmentof the radical (7–40 ms), a formation rate of 140 6 30 s21 wasestimated (Fig. 2f). This rate is as high as steady-state enzymaticturnover (150 s21) and therefore shows kinetic competence of theradical intermediate. Radical 4 is therefore a genuine intermediatebut has a low population under turnover conditions. Apparently theelectron is mainly located at the [4Fe–4S] cluster.

Support for the catalytic competence of the allylic ketyl radicalintermediate was gained by the effect of the substrate analogue2-hydroxypent-4-enoyl-CoA on the dehydratase. The molecule waspredicted to arrest catalysis because the resonance-stabilized penta-dienyl ketyl radical intermediate (Fig. 3, inset) would be too stable toundergo reoxidation to the product. Steady-state kinetics did indeedreveal that 2-hydroxypent-4-enoyl-CoA acted as a potent inhibitor,competitive with the natural substrate (Ki 5 62mM). Incubation withthe dehydratase gave rise to an EPR signal (Fig. 3) that, also on rapidfreeze-quenching, amounted to (12 6 1)% of the enzyme concentra-tion. The value of gav (2.0030 6 0.0002) and estimated isotropic pro-ton hyperfine couplings (1.11, 0.96 and 0.88 mT) match thoseexpected for the pentadienyl ketyl radical (Supplementary Table 5).In particular, the intermediacy of the 59-deoxyadenosyl radical inadenosylcobalamin-dependent and S-adenosylmethionine-dependentradical enzymes was demonstrated by a similar methodology withanhydroadenosyl derivatives20. In such systems, the stability of a crucialbut undetectable intermediate was increased as a result of resonance-stabilization of an allylic radical species.

Most of the radical enzymes described previously22 require eitherpreformed biradical oxygen or homolysis of covalent bonds in thespecial cofactors coenzyme B12 or S-adenosylmethionine for radicalgeneration. Here we have shown experimentally that under anaerobicconditions an enzyme can generate a radical intermediate with aniron–sulphur cluster as a prosthetic group and a single electron as acofactor. This mechanism could represent an ancient way of radicalformation preceding the evolution of other, more recent, types ofradical generator. The consumption of valuable proteinogenic aminoacids and the production of short-chain fatty acids seem to be anunlikely evolutionary driver of the 2-hydroxyacyl-CoA dehydratases.Rather, because dehydration of 2-hydroxyacyl-CoA is a reversiblereaction23, involvement in primordial amino acid biosynthesis from2-enoates seems a more likely possibility.

METHODS SUMMARY

All manipulations were performed under strictly anaerobic conditions in an

anaerobic glove box (Coy Laboratories). For handling outside the glove box,

the samples were protected from air by anaerobic manipulation by using syringes

with Teflon plunger tips and employing glass vials tightly stoppered with rubber

bungs. 2-Hydroxy-4-methylpentanoyl-CoA dehydratase was purified16 from C.

difficile (DSMZ 1296T) and its activator was obtained16 from Escherichia coli by

overexpression of the gene. After activation of dehydratase (66mM) for 5 min at

23 uC with an equimolar concentration of activator, the CoA ester (final con-

centration 1 mM) was added and the sample was frozen in a quartz tube. EPR

spectra were recorded on a Bruker EMX-6/1 spectrometer with a standard TE102

rectangular cavity, cooled by an Oxford Instruments ER-4112HV helium-flow

cryostat and simulated with the program SAE07 supplied by S. P. J. Albracht24.

For rapid freeze-quenching, activated dehydratase was mixed with an equal

volume of CoA ester in an SFM-20 instrument interfaced with MPS32 software

(BioLogic) and frozen in liquid-nitrogen-cooled isopentane. All reported errors

are s.e.m.

Full Methods and any associated references are available in the online version ofthe paper at www.nature.com/nature.

Received 6 August 2007; accepted 10 January 2008.

1. Buckel, W. in Biology of the Prokaryotes (eds Lengeler, J. W., Drews, G. & Schlegel,H. G.) 278–326 (Thieme, Stuttgart, 1999).

2. Buckel, W. & Keese, R. One electron redox reactions of CoASH esters in anaerobicbacteria. A mechanistic proposal. Angew. Chem. Int. Edn Engl. 34, 1502–1506(1995).

3. Buckel, W., Hetzel, M. & Kim, J. ATP-driven electron transfer in enzymatic radicalreactions. Curr. Opin. Chem. Biol. 8, 462–467 (2004).

4. Boll, M. & Fuchs, G. Unusual reactions involved in anaerobic metabolism ofphenolic compounds. Biol. Chem. 386, 989–997 (2005).

5. Seebach, D. Methods of reactivity Umpolung. Angew. Chem. Int. Edn Engl. 18,239–258 (1979).

6. Bruckner, R. Reaktionsmechanismen (Spektrum Akademischer, Heidelberg,2003).

7. Buckel, W. & Golding, B. T. Radical enzymes in anaerobes. Annu. Rev. Microbiol.60, 27–49 (2006).

331 334 337 340 343Magnetic field (mT)

H H

CoAS

O– H H

H

Figure 3 | EPR spectrum of the pentadienyl ketyl radical formed from the2-hydroxypent-4-enoyl-CoA substrate analogue. On incubation ofactivated 2-hydroxy-4-methylpentanoyl-CoA dehydratase with2-hydroxypent-4-enoyl-CoA, an intense radical EPR signal is generated(black trace). The spectrum can be simulated (red trace) with hyperfinesplitting parameters appropriate for the pentadienyl ketyl radical structure(inset). For experimental details and simulation see Methods.

NATURE | Vol 452 | 13 March 2008 LETTERS

241Nature Publishing Group©2008

8. Smith, D. M., Buckel, W. & Zipse, H. Deprotonation of enoxy radicals: theoreticalvalidation of a 50-year-old mechanistic proposal. Angew. Chem. Int. Edn Engl. 42,1867–1870 (2003).

9. Sebaihia, M. & Thomson, N. R. Colonic irritation. Nature Rev. Microbiol. 4,882–883 (2006).

10. Reineke, J. et al. Autocatalytic cleavage of Clostridium difficile toxin B. Nature 446,415–419 (2007).

11. Selmer, T. & Andrei, P. I. p-Hydroxyphenylacetate decarboxylase from Clostridiumdifficile. A novel glycyl radical enzyme catalysing the formation of p-cresol. Eur. J.Biochem. 268, 1363–1372 (2001).

12. Sebaihia, M. et al. The multidrug-resistant human pathogen Clostridium difficilehas a highly mobile, mosaic genome. Nature Genet. 38, 779–786 (2006).

13. Barker, H. A. Amino acid degradation by anaerobic bacteria. Annu. Rev. Biochem.50, 23–40 (1981).

14. Elsden, S. R. & Hilton, M. G. Volatile acid production from threonine, valine,leucine and isoleucine by clostridia. Arch. Microbiol. 117, 165–172 (1978).

15. Kim, J., Darley, D., Selmer, T. & Buckel, W. Characterization of (R)-2-hydroxyisocaproate dehydrogenase and a family III coenzyme A transferaseinvolved in reduction of L-leucine to isocaproate by Clostridium difficile. Appl.Environ. Microbiol. 72, 6062–6069 (2006).

16. Kim, J., Darley, D. & Buckel, W. 2-Hydroxyisocaproyl-CoA dehydratase and itsactivator from Clostridium difficile. FEBS J. 272, 550–561 (2005).

17. Hans, M., Buckel, W. & Bill, E. The iron–sulfur clusters in 2-hydroxyglutaryl-CoAdehydratase from Acidaminococcus fermentans. Biochemical and spectroscopicinvestigations. Eur. J. Biochem. 267, 7082–7093 (2000).

18. Weil, J. A. & Bolton, J. R. Electron Paramagnetic Resonance: Elementary Theory andPractical Applications 2nd edn (Wiley, Bognor Regis, 2007).

19. Wu, W. et al. Lysine 2,3-aminomutase and trans-4,5-dehydrolysine:characterization of an allylic analogue of a substrate-based radical in the catalyticmechanism. Biochemistry 39, 9561–9570 (2000).

20. Magnusson, O. T., Reed, G. H. & Frey, P. A. Characterization of an allylic analogueof the 59-deoxyadenosyl radical: an intermediate in the reaction of lysine 2,3-aminomutase. Biochemistry 40, 7773–7782 (2001).

21. Layer, G. et al. The substrate radical of Escherichia coli oxygen-independentcoproporphyrinogen III oxidase HemN. J. Biol. Chem. 281, 15727–15734 (2006).

22. Stubbe, J. & Van der Donk, W. A. Protein radicals in enzyme catalysis. Chem. Rev.98, 705–762 (1998).

23. Buckel, W. The reversible dehydration of (R)-2-hydroxyglutarate to (E)-glutaconate. Eur. J. Biochem. 106, 439–447 (1980).

24. Beinert, H. & Albracht, S. P. J. New insights, ideas and unanswered questionsconcerning iron–sulfur clusters in mitochondria. Biochim. Biophys. Acta 683,245–277 (1982).

Supplementary Information is linked to the online version of the paper atwww.nature.com/nature.

Acknowledgements We thank R. K. Thauer for the use of his EPR spectrometer,and V. Schunemann and M. Bennati for the use of their freeze-quench instruments.

Author Information Reprints and permissions information is available atwww.nature.com/reprints. Correspondence and requests for materials should beaddressed to A.J.P. (pierik@staff.uni-marburg.de).

LETTERS NATURE | Vol 452 | 13 March 2008

242Nature Publishing Group©2008

METHODSDehydratase activity measurements. The dehydratase was activated by incuba-

tion with excess activator for 5 min at 23 uC in 50 mM MOPS pH 7.0 (buffer A),

5 mM dithiothreitol, 5 mM MgCl2, 0.4 mM ATP and 0.1 mM dithionite (total

volume 0.5 ml). Activity was measured spectrophotometrically in the anaerobic

glove box, using the absorbance increase of 4-methylpent-2-enoyl-CoA

(De290 5 2.2 mM21 cm21). The reaction was initiated by 0.5 mM (R)-2-

hydroxy-4-methylpentanoyl-CoA. Fe/S contents of both components and the

kinetic parameters were, within experimental error, identical to those reported

previously16.

Preparation of EPR samples. Dehydratase in buffer A containing 2 mM sodium

dithionite (buffer B) was activated with 5 mM dithiothreitol, 5 mM MgCl2 and

5 mM ATP (total volume 0.25 ml) inside ilmasil-PN quartz EPR tubes (outside

diameter 4.7 mm, inside diameter 3.8 mm). A freshly dissolved and neutralized

CoA-ester solution (10 mM) in 0.025 ml of buffer B was added through Tefzel

tubing (outside diameter 1.8 mm, inside diameter 1.2 mm) linked to a disposable

tip on a micropipette. Mixing was performed by pushing the tubing up and

down. After capping, the sample was frozen outside the glove box in liquid-

nitrogen-cooled isopentane (about 130 K) for time points between 15 and 60 s.

Other samples were frozen in liquid nitrogen.

Rapid freeze-quenching. To remove traces of dioxygen, glass ram syringes (1 ml

with Teflon plungers) were kept in the anaerobic glove box, and tube lines of the

SFM-20 instrument were filled with 10 mM sodium dithionite in buffer A for

several hours. After syringes and tube lines had been washed with buffer B, the

first syringe was filled with 5 mM freshly dissolved and neutralized (R)-2-

hydroxy-4-methylpentanoyl-CoA in buffer B; it was then removed from the

glove box and mounted. At the same time, dehydratase (120 mM) was activated

for 3.5 min with an equimolar concentration of activator in buffer B containing

10 mM dithiothreitol, 10 mM MgCl2 and 10 mM ATP in the second syringe and

handled in the same way. A rapid procedure was essential because the dehydra-

tase retained full activity for only 5–10 min (ref. 16). Four samples were shot

within 1–2 min, with ageing times determined by the ram speed, ageing tubing

volume, and the 2-cm flight between nozzle and liquid-nitrogen-cooled isopen-

tane. For convenient ice particle collection, 50-ml disposable step dispenser tips

were employed. The narrow tapered end was shortened to 6 mm and connected

with rubber tubing to an EPR tube. These assemblies were filled and cooled with

liquid-nitrogen-cooled isopentane, stored in Dewar vessels. After removal with

adjustable pliers for sample collection, the assembly was rapidly returned. Ice

particles were then packed with a Teflon-tipped steel rod and EPR spectra were

recorded. For performance checks, 100mM equine skeletal muscle myoglobin in

100 mM potassium phosphate pH 7.2 was mixed aerobically with 5 mM sodium

azide in the same buffer25. The rate of high-spin to low-spin conversion as

detected by EPR spectroscopy in freeze-quenched samples was, within experi-

mental error, identical to the rate of absorbance increase at 580 nm as deter-

mined with a conventional Hi-Tech SF-61MX stopped-flow apparatus.

EPR spectroscopy. For determination of g values, the magnetic field near thesample position was measured by an EMX-032T Hall field probe. A correction of

0.33 mT was applied as calculated for the strong pitch standard (g 5 2.0028) and

the microwave frequency (ER-041-1161 counter). Spin concentrations were

calculated by double integration against a standard of 10 mM CuSO4 in 2 M

NaClO4 and 10 mM HCl. Radical concentrations in freeze-quenched samples

could not be accurately double-integrated owing to a low signal-to-noise ratio.

Concentrations were therefore determined by amplitude comparison with a

reference allylic ketyl radical EPR signal, which could be double-integrated.

EPR conditions were as follows: microwave frequency 9.45–9.47 GHz, modu-

lation frequency 100 kHz, temperature 10 K, modulation amplitude 0.6 mT

(1.25 mT for Fig. 2a, b), microwave power 8mW (20 mW for Fig. 2a, b).

EPR simulation. Intrinsic limitations were imposed by the linewidth and sym-

metrical nature of the radical signal. In addition, the signal-to-noise ratio did not

allow resolution enhancement26 of our experimental spectra. Therefore only iso-

tropic proton hyperfine coupling constants (Aiso 5 1.15 mT and 1.45 mT) and an

average g value (gav 5 2.0038 6 0.0002) could be extracted from the experimental

data by means of natural abundance and 2H-labelled substrates (A2H 5 A1

H/

6.514; see ref. 18). However, the slightly asymmetric experimental spectrum withthe 1-13C-labelled substrate yielded information on the g-anisotropy (gx 5 2.0049,

gy 5 2.0038 and gz 5 2.0027, all 60.0003) and required an anisotropic 13C hyper-

fine tensor (Axy < 0.4 mT, Az 5 3.7 6 0.5 mT) for successful simulation. After

extraction of this information, an anisotropic coupling for the proton at C-3

was used. Principal components had a fixed Ax:Ay:Az ratio of 3:1:2 (with

(Ax1Ay1Az)/3 5 1.15 mT), following common practice for allylic radicals20.

The hyperfine coupling of the proton at C-4 was kept isotropic

(Aiso 5 1.45 mT), similar to that observed in other systems19,20. Collinearity of

the anisotropic hyperfine tensors with the g tensor was assumed, because simula-

tions were not sensitive to variations in Euler angles. Thus, all spectra could be

simulated with a single set of 1H, 2H and 13C hyperfine coupling parameters and

an identical gaussian linewidth of 0.95 mT. From the negligible effect of deutera-

tion at C-2, it was estimated that the hyperfine constant was less than 0.2 mT. In

the absence of data on 2H-labelled and 13C-labelled 2-hydroxypent-4-enoyl-CoA-

derived radicals, the parameters obtained for the inhibitor were not as accurate as

those for the (R)-2-hydroxy-4-methylpentanoyl-CoA-derived radical. Isotropic

proton hyperfine coupling constants and g-tensor values were used to minimize

the number of variables. The best simulation was obtained with a gav value of2.0030 6 0.0002, a gaussian linewidth of 0.77 mT and hyperfine constants for

single protons of 1.11, 0.96 and 0.88 mT (with estimated errors of 0.05 mT).

Simulations with narrower linewidth and two additional protons with a hyperfine

constant less than 0.3 mT also yielded satisfactory simulations.

25. Tsai, A. L., Berka, V., Kulmacz, R. J., Wu, G. & Palmer, G. An improved samplepacking device for rapid freeze-trap electron paramagnetic resonancespectroscopy kinetic measurements. Anal. Biochem. 264, 165–171 (1998).

26. Ballinger, M. D., Reed, G. H. & Frey, P. A. An organic radical in the lysine 2,3-aminomutase reaction. Biochemistry 31, 949–953 (1992).

doi:10.1038/nature06637

Nature Publishing Group©2008