The Mechanism of Action of Ethanolamine Ammonia Lyase,

SUMMARY Ethanolamine ammonia lyase catalyzes the coenzyme B12-dependent conversion of ethanolamine to acetaldehyde and NH,. When the enzyme-coenzyme complex is frozen in liquid nitrogen during the act of catalysis, an electron spin resonance signal equivalent to about 0.04 spin per active site can be detected. By saturation experiments, the signal con-sists of two components, a broad derivative signal composed of two peaks (g = 2.34 and g = 2.08) assigned to the cobalt of BL2r, and a narrow signal (g = 2.007) assigned to a free radical. With [l , 1-D2]ethanolamine as substrate, the shape and position of the signal was unchanged, but its size was in-creased. In the absence of substrate, a signal was still observed, but the concentration of unpaired electrons was re-duced. The broad metal signal was unchanged except for a decrease in size. The radical signal was narrower (10 gauss, compared with 18 gauss in the presence of substrate) and was shifted (g = 2.003) with respect to the substrate-dependent signal, suggesting that the radical in the absence of substrate may have been different from the substrate-dependent radical. These observations support the proposal that the mechanism of action of this coenzyme Bly-dependent enzyme involves homolysis of the carbon-cobalt bond of the coenzyme.


SUMMARY
Ethanolamine ammonia lyase catalyzes the coenzyme B12dependent conversion of ethanolamine to acetaldehyde and NH,.
When the enzyme-coenzyme complex is frozen in liquid nitrogen during the act of catalysis, an electron spin resonance signal equivalent to about 0.04 spin per active site can be detected.
By saturation experiments, the signal consists of two components, a broad derivative signal composed of two peaks (g = 2.34 and g = 2.08) assigned to the cobalt of BL2r, and a narrow signal (g = 2.007) assigned to a free radical.
With [l , 1-D2]ethanolamine as substrate, the shape and position of the signal was unchanged, but its size was increased.
In the absence of substrate, a signal was still observed, but the concentration of unpaired electrons was reduced.
The broad metal signal was unchanged except for a decrease in size.
The radical signal was narrower (10 gauss, compared with 18 gauss in the presence of substrate) and was shifted (g = 2.003) with respect to the substrate-dependent signal, suggesting that the radical in the absence of substrate may have been different from the substrate-dependent radical. These observations support the proposal that the mechanism of action of this coenzyme Bly-dependent enzyme involves homolysis of the carbon-cobalt bond of the coenzyme.
Investigations of the mechanism of action of t.his enzyme have indicated that. the initial step in this reaction involves cleavage of the cnarbon-cobalt bond of the coenzyme, followed by the transfer of hydrogen from the substrate to the adenosyl group of t,hc caoenzyme to form 5'.deoxyadenosine (1). Rearrangement then takes place, followed by the re- (1) as well as esperiments using spin-labeled cobamide coenzyme (3) both point to this conclusion.
The first evidence, howcvcr, was the demonstration t,hat upon addition of .qubstrabe to enzyme-coenzyme complex, an electron spirl resonance signal appears (4). The instrumentation used for that study was 1101 adequate for a close analysis of the properties of the signal, which was llresent at very low levels.
The availability of better equipment IIOW llas permitted the signal to be studied in more dctnil. This st,udy is the subject of the present report.

MATERIALS
AXI) HETHOl)S Etlianolamiile aninloilia lyase from Cloalridiwt~ ~1). ww l)rcpared and resolved of lmmd cobamides by the mrthod of Kaplan and Stadtman (5). Enzyme concentration was calculated on the basis of a molecular weight of 520,000 (6). The enzyme had previously been shown to possess two active sites per molecule (7). The turnover munher of t'he enzyme I)reparation used in these experiments was 7,600 min' at 23", using lo-" M coenzyrnc. Coenzyme B12 was purchased from Calbiochem.
Isotopically labeled ethanolamine was prepared by the reduction of glycine ethyl ester by LiAlD4 (Alfa Inorganics) as previously described (8). Incubations were conducted at 23" in quartz ESR' tubes. The reaction was begun with enzyme which had been allowed to stand at, room temperature for 10 min. ht the appropriate time, the incubation was terminated by plunging t,he ESR tube into a container of isopentane immersed in liquid nilropen (77" K). The frozen sample was t,hen t,ransferred to the cavity of a Va.ri:tn E-9 electron spin resonance spectrometer for spectroscopy. All operations were conducted in dim light to I)revent photolysi~ of the coenzyme. 1. l<Sl1 spcctnml 01 the enzyme-U>IBC complex in the presence and absence of subst.rate. The reaction mixtures coottaincd 15.4 nmoles of ethanolamine ammonia lyase, 44.7 nmoles of I)MBC, 15 rmoles of ethanolamine. NC1 (pH 7.4)) and 2 ~moles of potassium phosphate buffer (~1-1 7.4), with omissions as noted, iI1 a volume of 0.3 ml. Incubations (45 s) were conducted as described in the text,. The 15Sli spectra were taken at, 9.5" K. The microwave l'reqliency was 9.1*58 (iHa, and the micro\vavc po\vcl was 1.0 milliwatts. Modulatioll amplitude (10 gallss) atid gait1 werr! the same for bot.h spcct rti. 2. Rate of appearance of the substrate-dcpclltterit, 1SSIt signal. The reaction mixtures contained 7.0 nmoleu of etheuolamine ammonia lyasc, 44.7 nmoles of DMBC, 1.j ~mulcs of cthanolamine. HCI (pH 7.4), and 2 pmolcs of potassilnn phosphate bul'fer (pH 7.4) in a volume of 0.3 ml. Incnbatioirs were condnctcd fat the times noted as described in the text. J5Sl< spectra were taken at 9,5" K. The conditions were as described tlllder Fig. 1 except that the microwave power was 10 milliwatts.
(;ain \vits the same for all spectra. The relative amplit.ude 1, as est imatcd 1)~ measuring the dist,ttnre betwccll the peak and the trorlph of the mH,jor signal.
the incubation and then subjected to ES11 hl)cctroln(~try ,zt liquid nitrogen temperatures, :I well defined 11:~row signal was observed. This signal (showu in Fig. 1, fop) was lot&et1 at u -X007* and had a peak to peak width of 18 gauss. iroyjerties suggest that it was generated by :I fret radical. 'l'hcse A signal was also observed iI1 the absence of srlbstratr (Fig. 1, hottom). This signal was much smaller than 11-1~ one generated in the presence of substrate.
From its width and location, it too appears to represent, a free radical.
The rate 0T atltmwiulce of the .~;rll~sti~wtf~'-dcI~~~iflf~iit sigld is shower ill Fig. 2 Owr tlic following 15 > t,herc sclrrnrd t,o be a slow illcrease in the size of the signal. I IORrver, the difference betwern the 15-s value and the 60-s v:duc was close enough IO experimental error that 21 conclusion cannot hr drawl) about whelher this inrrrasr ill sigtlal size :tct~~all> rc1)rc~seiit.s x clisligc in thtx c~otlcciitratioil of frer r:itlic:d or not When the spectrum was sci~med over :I wider range of firltl st.rcngtlls, it became apparent that there wel'r two OI her sigll:A ln'rscnt iii addition to the signal geiirratrtl by the I'rt'c r:ltlical. These signals ilr(l shown iii Pig. 3. In oidrr to brii~g out tlrcii fr:rt,urc~s, tllis part~icular spcctrunr \vas takrll utrder coirditiotls iii which the radical signal waz estcnhivel\-satui~atcd. 'I'hr \VWk field signal is locaatrd at g = 2.34, and 111~ strong firid sigll:d :t t g = 2.08. 'J'hcy are both broiad, with peak to peak widths of 90 gauss and about 65 gauss, resl)cc~tively. \i'itll rehllect to botll their posit,ion and their width, theqe signals are rcminisr(~llt 01 those generated by the paramagnetic d7 cobalt atom in H1?,. (9, IO). 111 order to obtain evidence 21,s to bow many chemical sl)Pries were responsible for generating these thrre signals, saturation csperimrnts were performed.
Iii llie filxst of these esperirncxnts, a conil)arison was made of the anrl)litucles of ihe "r:L~lic:il" l)& :mtl the broad peak at g = 2.08 :IS a function of micro\vuve l)ow~i delivered t,o the cavity of the spectrometer.
The results of this experiment, presented in Fig. 4, show th:it the "radi(~nl" peak is severely ,sat.urated at lcvcls of power well below those rrquired to saturate the broad peak. This observation cw~kirnrs that 1 he two signals are generated by different species and supports the idea that the narrow signal represents a radical and the broad signal rel)rcsrnts a 1)aramagnetic metal ion. 011 the other hand, the ratio of amplitudes of t.he t\vo broad pc:lkh is constant over R widr range of temperatures and l)ower Irvels (  4. Amplitudes of the g = 2.007 and g = 2.08 sigrlals as a function of power. Thr reaction mixture wets the same one t~sed for the substrate-depeadent spectrum in Fig. 1. JMlZ spectra were taken at 95" K. Except for the microwave power, which varied as shown in the figure, the conditions were as described under Fig. 1. The relative amplitude was estimated by dividing the peak-to-trough distance of the signal in question by t,he gain. The amplitude of the signal at 1 milliwatt was arbitrarily set at 1.0.

T.WLE I
Avlpliludes of g = 6.08 and q = 2.%$ signals US fun,clion of power and temperalure The reaction mixture was the same one used for the substratcdependent spectrum in Fig. 1. The 15SR spectra were t&en at various temperatures and power levels as shown in t.he table. Modulation amplitude was 10 gauss for all spect,rrt. llelative amplitudes of the peaks in question were determined as described in Fig. 4. The amplitude of the g = 2.34 peak at 95" K and 1 milliwatt, power was arbitrarily set at) 1.0. the same paramagnetic species. The fact that there are two Ileaks suggests that this species is at least approsimabely in an axially symmetrical environment.

Conditions
This result provides furthrl support for the assigrlment of t.hese signals to t,hc divalent cobalt of r31nr.
The concentration of unpaired electrons ill the reaction mixture was determined by double integration of the spectrum with 1.0 InM Cu++-EDTA as standard, Ry this technique, ljaramagnetic cobalt was estimated to be present at a concentration of 0.04 g atom per mole of active sites. It was not possible to determine the true concentration of free radical, since even at the lowest powers obtainable the signal was extcltsively saturat,ed at 95" K (see Fig. 4). However, by double integration it was determined that the radical concentration substantially exceeded 0.01 mole per mole of active sites.
Incubations were also done with [I, l-Da]etilarlolanlille as substrate.
One incubation was conducted for 45 s, just, as with unlabeled substrate.
However, a 45-s incubation with deuterium-labeled substrate is not strictly comparable to a 45-s in-cubaCon with unlabeled substrate, since a kinetic isotope effect of 7 is observed with the labeled material (II).
With deutrrated substrate tllere was a significant, increase in the amplitude of both the radical and the metal signal5 (Table  II).
This was SCCII iu both the 45-s :uld 5.min incubatioils. It indicates that the collcacntratioll of ulll)aired electrons gcllcratrd in the presence of dcuterat,ed ethanolamine is greater than the concrlltlation ill thcl presence ol' u~~lnl~lcd subst,ratr.

DISCUSSION
The first suggestion that coeuzymc l~,a-deprndcnt rcactiolls might involve intermediates with unpaired electrons was made by Eggerer et al. (12) on the basis of studies on the (soenzyme Bin-dependent rearrangement of methylmalonyl-CoA. Direct evidence supporting this hypothesis only became available with the observation that ESR signals appear in reaction mixtures containing ethanolamine ammonia lyase, coenzyme 1112, and subst,rate (4). Since this observation was made, much evidence has accunlulated indicating that homolysis of the carboll-cobalt bond ta.kes plare during the ethanolamine ammonia Iyrrse reaction (see above). In addition, recent. evidence has indicated that other coenzyme 1j12-dependent enzymes may rupture the carbon-cobalt bond of the coenzyme homolytically. There are several papers concerning ESR signals generated by coenzyme B12-dependent ribonucleotide reductase under various circumstances (13,14). Most recently, reports have appeared documenting the appearance of ESR signals during catalysis by diol dehydrase and glycerol dehydrase, both coenzyme R~dependent enzymes which catalyze the coIlversion of glycols to aldehydes (15,16).
If catalysis by ethanolamine amnlonia lyase involves homolysis of t,he carbon-cobalt bond of the coensyme, it would be expected that an ESR signal generated by this process should show a component due to the paramagnetic cobalt atom at~d one due to an unpaired electron on a free radical.
Although previously obtained ESR spectra showed only a metal signal (4), both of these components were observed in the present st#udy. Failure to observe the free radical component in the previous ESR study was probably due to a combination of instrumental inadequacy and saturation of the radical signal at the power levels used.
An additional cobalamin species has been described which can be observed by ESR spectroscopy, viz. peroxocobalamin (17). It is unlikely that, either of the signals reported here represents this species. The g values for the metal signal seen in the present With regard to the radical signal, not only is its g value in rather poor agreement with either of the peroxocobalamin g values, but its width is only about onefifth that of the signal produced by peroxocobalamin.
From the properties of the signal, it appears that a steady state situation exists involving enzyme-coenzyme complex with an intact carbon-cobalt bond (diamagnetic) and complex with a dissociated bond (paramagnetic).
The fact that an ESR signal is present in the absence of substrate indicates that under these conditions the carbon-cobalt bond is dissociated to a small extent.
In the presence of substrate the signal undergoes alterations indicating a shift in the steady state in favor of the species in which the bond is dissociated as well as a probable change in the identity of the radical. That the situation in the presence of substrate is a steady state (or quasi-steady stat)e), involving complexes with intact and broken carbon-cobalt bonds, is indicated by the time course (Fig. 2), which shows that the concentration of unpaired electrons rises rapidly within the first 15 s or less to a steady state level which then either remains constant or rises slowly over the next 45 s. The steady state level of carbon-cobalt bond cleavage of 470, calculated from the concentration of unpaired electrons on the cobalt atom, agrees with previous experiments in which the steady state level of carboncobalt bond cleavage was measured direct'ly with W-labeled coenzyme Bls (1).
The ethanolamine ammonia lyase reaction displays a significant lag during its early stages, the maximum reaction rate not being achieved until 30 set to 1 min after t,he reaction is begun (5). The present experiments show that the rate at which the concentration of unpaired electrons approaches the st'eady state is at least as rapid as the rate at which the deamination of ethanolamine approaches its final velocity. This result supports the concept that the paramagnetic species generated in the presence of substrate are involved in catalysis, since they appear at a rate which is kinetically competent with respect to the cat.alytic reaction.
Previous work has shown that there is almost total disappearance of the signal from a spin-labeled coenayme Blz during catalysis by ethanolamiue ammonia lyase (3), indicating virtually quantitative cleavage of the carbon-cobalt bond in the steady state, while the present experiments show only 4 to 5yo carboncobalt bond cleavage in the steady state when ordinary coenzyme is used. This difference may be due to differences in the conditions used in the two sets of experiments, to intrinsic differences in the ease with which the enzyme is able to cleave the carboncobalt bond of the two species of coenzyme, or perhaps to differences in the rates of the various individual steps of the overall reaction with the two species of coenzyme.
In the ethanolamine ammonia lyase reaction, the rate-limiting step, involving a reversible transfer of hydrogen between 5'deoxyadenosine and product, is preceded by a rapid irreversible step (Scheme 1) (11 This finding suggests that the cleavage of the carbon-cobalt bond in the presence of substrate is directly rollnetted to the hydrogen transfer step and is not merely relat'ed to the binding of substrate to the enzyme. In most of the experiments, the conditions were such tllat the substrate had been completely consumed by the time the reaction was terminated by freezing in liquid nitrogen. The ESR signals in these experiments must therefore be due to paramagnetic species involved in the ammonia-dependent exchange of hydrogen between acetaldehyde and coenzyme B,*, a process which constitutes a partial reversal of the ethanolamine ammonia lyase reaction (2). The single experiment in which the deamination of ethanolamine was still taking place when the reaction mixture was frozen was the 45-s incubation with deuterium-labeled substrate.
The ESR signal from this incubation was virtually identical in configuration and amplitude with the signal from the 5-min incubation with deuterated substrate, indicating that the same paramagnetic species are produced during the deamination of ethanolamine as during the acetaldehyde-coenzyme Blz hydrogen exchange. This result provides further evidence supporting the formulation shown in Scheme I.