Evidence That the Greening Ligand in Native Butyryl-CoA Dehydrogenase Is a CoA Persulfide*

Yellow butyryl-CoA dehydrogenase and general acyl-CoA dehydrogenase are “greened” by a mixture of coenzyme A plus elemental sulfur. The resultant stable complex contains an identical ligand with that present in native green butyryl-CoA dehydrogenase and has the same broad absorption band centered at 710 nm. Evidence is presented that the greening ligand is a CoA persulfide, possibly a mimic of the substrate carbanion thought to be generated early in the normal catalytic cycle. Variation in the position of the long wavelength band on replacement of FAD by a series of analogs of differing oxidation-reduction potential is consistent with a charge-transfer complex between a persulfide as the donor and oxidized flavin as the acceptor. The possible physiological and metabolic significance is dis-cussed. Acyl-CoA dehydrogenases mammals catalyze the f i t step in fatty acid oxidation, with the introduction of a trans double bond between C-2 and C-3 of their substrates.

saturated water. The specific activity of this solution was 1.82 X IO5 Bq nmol". Corrections for the decay of %S were applied as recommended by the manufacturers. Both labeled solutions were stored at -20 "C. The sodium sulfide was kept under nitrogen and in the dark. A quench curve for each was constructed using the sample channels ratio method (16). 14C and %S were estimated independently in a mixture of xy1ene:Triton X-100:2,5-diphenyloxazole (700:300:5, v/v/ w) (16), using the I4C window of a Philips PW 45400 liquid scintillation analyzer.
Materials-Coenzyme A was purchased from Boehringer Mannheim. Butyryl-CoA was prepared from CoASH and butyric anhydride (17). Acetoacetyl-CoA was prepared from CoASH and freshly vacuum-distilled diketene (18,19) and assayed in MgS04 (20). Colloidal sulfur was produced from acidified sodium thiosulfate (21), stored in water, and vigorously shaken before use. Other chemicals were of the highest grade commercially available.
Persulfide Estimation-Samples were incubated for at least 30 min after addition of KCN to a final concentration of 5 mM. The thiocyanate produced was estimated by 460 nm absorbance after addition of "Sorbo's reagent" (22,231 using a standard of ammonium thiocyanate. For this assay, E at 460 nm = 4.0 m"' cm". Taurine Analyses-Amino acid analyses with ninhydrin detection were performed on a single column Locarte bench model amino acid analyzer, interfaced to an MP6800-based integration computer. Samples were first dialyzed against 1 mM KPi, pH 7, and then hydrolyzed in 6 M HCl and 5% dimethyl sulfoxide (24) at 110 "C for 24 h. Under these conditions, CoA breaks down to given stoichiometric amounts of taurine and p-alanine (3). The exact taurine content for each sample was estimated in two consecutive analyzer runs, one (10 nmol) to allow accurate estimation of taurine and the second (1 nmol) for estimation of the other amino acids. The taurine/proline ratio served as a convenient estimation of the protein CoA content expressed in all cases per mol of enzyme sybunit.
Reconstitution of Apoprotein with Flavin Analogs-The apoprotein of general acyl-CoA dehydrogenase was prepared by the method of Mayer and Thorpe (25). Reconstitutions were performed at 4 "C or 25 "C for 1-2 h in 50 mM KPi, pH 7.6, using 25-36 p~ apoprotein subunits and 1-2 eq of flavin analog. Unbound flavin was removed by gel filtration or dialysis against 50 m~ KPi, pH 7. Details of the conditions will be published elsewhere. Fig. 1 shows the spectrum of yellow butyryl-CoA dehydrogenase and the complex produced after aerobic incubation of the yellow form with CoASH and Na2S. The latter spectrum is indistinguishable from the published spectrum of native green butyryl-CoA dehydrogenase (3). Various forms of sulfur when added to yellow butyryl-CoA dehydrogenase together with CoASH are found to "green" the enzyme at markedly different rates with no change in activity in the dye-coupled catalytic assay. Qualitatively, the slowest greening is seen when Na2S is added to a mixture of CoASH and yellow enzyme; after 15 min, the A71O:A430 ratio reaches 0.43. However, at equivalent concentrations, if Na2S and CoASH are first incubated together for 15 min, before adding to the enzyme, the A710:A430 ratio reaches 0.43 after 2 min. This strongly implies a reaction between the CoASH and S2before binding to the enzyme. Flowers of sulfur (a So species which exists as SS rings) dissolve in a solution of NazS to give a mixture of linear polysulfanes (21)

RESULTS AND DISCUSSION
where n = 2,3,4 etc., dependent on the concentration of S2-. If an aliquot of this mixture is added to yellow butyryl-CoA dehydrogenase and CoASH, relatively rapid greening occurs; the A710:A430 ratiobreaches 0.48 after 5 min. Furthermore, a suspension of colloidal sulfur in water, when added to yellow butyryl-CoA dehydrogenase and CoASH gives an A710:A430 of 0.37 after 5 min. An SO species of some sort is therefore clearly implicated in the greening reaction.
In control experiments, 1 mM Na2S was incubated with two different samples of yellow butyryl-CoA dehydrogenase without any added CoASH. Unexpectedly, the spectrum of one sample after incubation showed an A710:A430 ratio of 0.28 (52% green). Amino acid analysis showed, however, that some endogenous CoA remained bound even after extensive anaerobic dialysis and that this sample of enzyme still contained 0.60 mol of CoA/mol of enzyme subunit. The second sample gave an A710zA430 ratio of 0.13 (24% green) with 0.30 mol of CoA bound/mol of enzyme subunit. These data indicate that CoASH is a prerequisite for greening and that approximately 1.2 mol of CoA/mol of enzyme subunit would be required to produce fully green enzyme (see later).
Similar experiments with general acyl-CoA dehydrogenase (which contains no bound CoA after purification (12)) showed slow greening when Na2S was incubated with a mixture of CoASH and the enzyme. As with butyryl-CoA dehydrogenase, rapid greening was observed after prior incubation of CoASH and NaZS, or when the polysulfane solution replaced the Na2S. The maximum A710:A430 ratio achieved for general acyl-CoA dehydrogenase was only 0.22. Furthermore, this complex is less stable than that of butyryl-CoA dehydrogenase; the A710:A430 ratio falls to one-half after 20 h in aerobic 100 mM KPi buffer, pH 7.6, at 25 "C. Native general acyl-CoA dehydrogenase, with no bound or added CoASH, showed no greening with Na2S alone.
Persulfide Formation in Free Solution-The postulated reaction between CoASH and Na2S in free solution was investigated further. In the absence of enzyme, 10 mM Na2S and 0.48 mM CoASH were incubated in 0.2 M NaOH. This was undertaken to observe spectral changes in the 330-340 nm region, indicative of persulfide formation (26). The increase of 0.044 at 335 nm after 3 h is equivalent to 0.14 m~ persulfide ( E at 335 nm = 300 M" cm", see Refs. 26 and 27). Lowering the pH to e 3 completely abolished the 335 nm absorption band, again consistent with a persulfide (26). These changes are also seen to a lesser extent in 100 mM KPi, pH 7, i.e. under the conditions of the greening experiments described above.
Correlation between f14C]CoASH Binding and Greenness- Fig. 2 shows a close correlation between increasing greenness and the amount of [14C]CoA incorporated. However, the sample of yellow butyryl-CoA dehydrogenase used, when incubated with Na2S alone, gave an A710:A430 ratio of 0.17. This is due, as mentioned above, to endogenous bound CoA which is difficult to remove entirely. This complicates the interpretation of these results, as it is unclear to what extent the added radioactive CoA exchanges with the bound shown to produce no degreening (3). Therefore, the [35S] greened butyryl-CoA dehydrogenase was incubated for two consecutive periods of 48 h in 5 mM KCN at 4 "C (Table I). A more rapid decline in bound =S was seen with no significant decrease in greenness. Analysis of fractions collected from a calibrated Sephadex G-25 column (Fig. 3) indicates release of both 35S2and CN35S-. The nongreening 35S is therefore in the form of both sulfide and cyanolyzable sulfur. The resulting stoichiometry after two KCN treatments (Table I)  Stoichiometry of 3sS Binding-Sixty-four nmol of yellow butyryl-CoA dehydrogenase, 1740 nmol of Na2S, and varying amounts of CoA up to 192 nmol were incubated until greening reached completion. Unbound 35S was removed by the treatment described in Fig. 2 for the removal of unbound ["C] CoASH. Unlike the incorporation of ['4C]CoASH, however, the amount of 35S initially bound showed a poor correlation with greenness. The stoichiometry of 35S incorporation into the greened samples, when extrapolated to fully green enzyme, ranged from 1.2 to 2.7 mol of 35S/mol of butyryl-CoA dehydrogenase. For example, one butyryl-CoA dehydrogenase sample was 68% green and contained 1.72 mol of 35S/mol of enzyme subunit (2.5 mol of "S/mol of green butyryl-CoA dehydrogenase). Another sample, treated in the original incubation with a different amount of CoASH, was 92% green and contained 1.19 mol of %/mol of enzyme subunit (1.3 mol of 35S/mol of green butyryl-CoA dehydrogenase). These seemingly inconsistent results suggested that some nongreening 35S was bound to butyryl-CoA dehydrogenase, possibly as sulfide or as persulfide. In keeping with this view, it was found that, when [35S]greened enzyme was simply stored for 10 days at 4 "C in buffer at pH 7, a pronounced decline in the bound 35S occurred with little or no concomitant decrease in greenness ( Table I). The nongreening, relatively weakly bound sulfide is presumably in equilibrium with free sulfide in solution which is then slowly lost to the air as gaseous H2S. Additionally, persulfides are normally unstable and readily form sulfide. After storage, the stoichiometry of residual 35S bound to the enzyme was close to 1 mol/mol of fully green butyryl-CoA dehydrogenase (Table I).
Cyanide reacts with persulfides to give free thiocyanate (22, 23), so that if some nongreening 35S is present as persulfide, then incubation in KCN might be expected to accelerate the removal of bound 35S (to given free CNS-). Incubation of native butyryl-CoA dehydrogenase with 5 mM KCN has been Ganidinium chloride at pH 7, (iii) NaOH added to pH 10, or (iv) 5 mM KCN and NaOH to pH 10. Samples were then separately applied to the calibrated column. Examination of collected fractions showed a 260 nm peak of denatured protein at 14 ml and a 450 nm peak of released FAD at 35 ml. 35S was distributed between the protein peak and the CoASH/S2peak. In the case of (iv), 35S was also seen under the thiocyanate peak. In considering the results from Table 11, it is important to bear in mind that although this enzyme sample is 74% green, it contains 1.00 mol of 35S/mol of butyryl-CoA  The conversion of cyanolyzable sulfur from the CoASH peak i n (iii) to the CNS-peak in (iv) can be clearly seen.
dehydrogenase. On the assumption that the greening ligand contains only one extra S atom in addition to that already present in CoA, this sample presumably contains 0.26 mol of nongreening 35S/mol of butyryl-CoA dehydrogenase. In view of this, one may attempt to account for the results in Table 11.
(a) Both CoASH and S2-(and therefore presumably CoA-S-S-) elute at very close positions from the calibrated Sephadex column, and hence it is not possible to distinguish directly between a CoA persulfide and free S2-. However, an estimation of the CoA persulfide/S2-ratio under this peak is possible by comparing (iii) and (iv), since denaturation in KCN converts all cyanolyzable sulfur to CNS-which will elute under a separate peak. 0.87 mol of 35S/mol of butyryl-CoA dehydrogenase loaded is found under the CoASH/S2peak in the case of alkaline denaturation (iii). The addition of KCN to the denaturing solution (iv) results in only 0.12 mol of 35S/mol of butyryl-CoA dehydrogenase loaded remaining under the CoASH/S2-peak; thus, 0.75 mol of 35S/mol of butyryl-CoA dehydrogenase loaded has been transferred from the CoASH/S2-peak to the CNS-peak due to cyanolysis of the postulated CoA persulfide. The hypothesis of 1 mol of "S"/mol of butyryl-CoA dehydrogenase for fully green enzyme would require for this sample, which is 74% green, that 0.74 mol of "S"/mol of butyryl-CoA dehydrogenase existed as persulfide associated with CoA.
(b) For (i) to (Si), the quantity of 35S under the denatured protein peak varies from 0.13 to 0.23 mol of "S/mol of butyryl-CoA dehydrogenase loaded. Most of this exists as cyanolyzable sulfur, since denaturation in KCN, which will convert proteinpersulfides to CNS-, leaves only 0.04 mol of 35S/mol of butyryl-CoA dehydrogenase loaded under the denatured protein peak. The protein-bound cyanolyzable sulfur may be totally nongreening 35S.
(c) Denaturation in KCN and NaOH (iv) shows, by counts under the CN35S-peak, that the total content of cyanolyzable sulfur before denaturation is 0.84 mol of 35S/m~l of butyryl-CoA dehydrogenase. T h i s is in close agreement with the assay for thiocyanate in the collected fractions using "Sorbo's reagent," which gives 0.80 mol of "S"/mol of butpyl-CoA dehydrogenase loaded.
(d) The total recovery of 35S improved as the pH increased. This is consistent with the destruction of persulfides to give free sulfide and subsequent partial loss to the atmosphere as the pH is lowered.
Various samples of native butyryl-CoA dehydrogenase with different A71~:A430 ratios were denatured in either 5 mM KCN with NaOH added to pH 10, or in 150 m KCN, and then applied to a calibrated Sephadex G-25 column (Fig. 3). The collected fractions were then assayed for thiocyanate as before. The samples showed a content of cyanolyzable sulfur ranging from 1.3 to 2.5 mol of "S"/mol of enzyme subunit extrapolated to fully green enzyme. Native butyryl-CoA dehydrogenase, therefore, contains cyanolyzable sulfur, some of which is not associated with greenness. Samples of the yellow form of the enzyme denatured in 150 mM KCN showed less than 0.15 mol of cyanolyzable sulfur per mol of enzyme subunit.
These denaturation experiments strengthen the supposition that the greening ligand is a compound containing a CoA moiety and one cyanolyzable sulfur atom. It appears that the enzyme also contains variable quantities of sulfur species which are not associated with greening. The cyanolyzable component of this additional sulfur could be attached to cysteine residues in persulfide linkage. Conceivably there also may be significant formation of coenzyme A polysulfides.

Binding of 35S to Bovine Serum Albumin and Pig Heart
Lactate Dehydrogenase-To examine the possibility that butyryl-CoA dehydrogenase might not be unique in binding sulfides "nonspecifically," separate solutions of bovine serum albumin and lactate dehydrogenase (7 mg d-') were incu-  Treatment of Butyryl-CoA Dehydrogenase with Dithionite-Reducing native butyryl-CoA dehydrogenase with dithionite abolishes the 710 nm absorption band. A sample of [35S]greened enzyme was treated with dithionite, followed by dialysis against 10 mM dithionite at pH 7. Subsequent oxidation of the enzyme yielded the yellow form of the enzyme with only 5% of the original 35S remaining bound. If the sample was reoxidized without dialysis, and then dialysed aerobically, 13% of the original 35S remained unbound. While dithionite reduction followed by anaerobic dialysis removes some bound CoA from butyryl-CoA dehydrogenase, dithionite reduction followed by aerobic dialysis does not (3). These data, therefore, show that degreening by dithionite is not due to loss of bound CoA, but to removal of the cyanolyzable sulfur atoms of the greening ligand, possibly by sulfitolysis.
Replacement of the Greening Ligand-The addition of acetoacetyl-CoA to yellow butyryl-CoA dehydrogenase instantaneously produces a complex with a broad absorption band centered at 580 nm (7). An excess of acetoacetyl-CoA, when added to the green form of the enzyme, also gives a 580 nm band after a few seconds.' Replacement of tightly bound ligand by an excess of another ligand in the acyl-CoA dehydrogenases has been well documented (7,28). In the displacement experiments reported here, the procedure was as follows. A 10-fold excess of acetoacetyl-CoA was added to [35S]greened enzyme and the decline in 710 nm and rise in 580 nm were monitored. When each was constant, the sample was applied to a Sephadex G-25 column and fractions containing the enzyme were pooled. These fractions were counted for radioactivity and spectra were recorded. A reduction in bound [ 14C] CoASH proportional to the decrease in A710:A430 and concomitant rise in A58o:A430 clearly showed replacement of the CoA moiety of the greening ligand. However, 35S was only partially displaced after binding of acetoacetyl-CoA. A reduction in the A710:A430 ratio of 0.31 (corresponding to displacement of 0.58 mol of the greening ligand per mol of enzyme subunit) was accompanied by a decrease in the bound 35S of only 0.30 mol/ mol of enzyme subunit. While the CoA moiety is clearly fully replaced, part of the cyanolyzable sulfur appears to remain. A possible explanation is thzt once the CoA persulfide is released from the enzyme, the cyanolyzable sulfur atom can migrate from the CoA sulfur to form similar persulfide links with cysteine residues by a simple exchange.
Stability of the Greening Ligand-Persulfides in free solution are unstable and at low pH readily break down to give sulfides (21). However, it is apparent that the green butyryl-CoA dehydrogenase complex is very stable-the greening ligand remains bound even after cold storage for several years, extensive dialysis, ammonium sulfate precipitation, and treatment in 3.5 M urea (3, 7, and Footnote 1). On the binding site, therefore, the persulfide linkage must somehow be stabilized. It is known that butyryl-CoA dehydrogenases bind CoA compounds extraordinarily tightly (28) and that these ligands can be removed only by denaturation (28), addition of an excess of competing ligand (28), or anaerobic dialysis of the reduced enzyme (3). However, degreening of butyryl-CoA dehydrogenase can be brought about by dithionite reduction without dialysis (3, 28) or by incubation in 160 mM sulfite.' After removal of the degreening agent by dialysis, regreening can be achieved only by the addition of a suitable sulfur species. On the other hand, butyryl-CoA dehydrogenase which has been degreened by phenylmercuric acetate can be regreened simply by addition of a thiol, such as mercaptoethanol or dithiothreitol (3, 9). No change in catalytic activity is G . Williamson The CoA-S-S-Hg-0 would remain bound (9). The "regreenability" (by thiols) of mercurial-treated enzyme declines with time and is more rapid at lower pH (9), which is consistent with the increasing loss of the persulfide sulfur into solution as the pH decreases. Addition of a thiol, however, would remove the phenyl mercury moiety to leave the CoA persulfide and thereby regreen. (ii) Green butyryl-CoA dehydrogenase may also be degreened by direct attack on the cyanolyzable sulfur atom. Sulfite attacks persulfdes or polysulfides even more readily than cyanide. This sulfitolysis yields thiosulfate, in a reaction analogous to cyanolysis (21). The mode of dithionite degreening may follow a similar mechanism. Dithionite is required at lower concentrations than sulfite to degreen, presumably because ligands are less tightly bound (and hence more accessible) when the FAD is reduced (3).
An apparent anomaly is the inability of 5 mM KCN to degreen native butyryl-CoA dehydrogenase (3). However, the CoA persulfide is shown to become accessible to 5 lll~ KCN after denaturation. Furthermore, green general acyl-CoA dehydrogenase, which is less stable than green butyryl-CoA dehydrogenase, is degreened by 5 mM KCN without concurrent denaturation. In the case of bacterial butyryl-CoA dehydrogenase, therefore, the ligand is presumably shielded against attack by its very tight association with the flavoprotein.
Several lines of evidence suggest that acyl-CoA derivatives are activated by removal of an a-proton prior to the flavin reduction step (29-32). Possibly, the persulfide mimics the transient carbanion by introducing a suitably orientated negative charge within the active site of the acyl-CoA dehydrogenases: Charge-Transfer Nature of the 71 0 nm Absorption Band-The suggestion that the 710 nm band is due to a chargetransfer complex (7) has been tested by using general acyl-CoA dehydrogenase reconstituted with several flavin analogs CoA Persulfide in Native Butyryl-CoA Dehydrogenase 4319 (see "Experimental Procedures"). The results are shown in Table 111, which lists the energy of the long wavelength band (expressed as VCT, recipro4 of the wavelength maximum) and known oxidation-reduction potentials of the flavin analog.
The charge-transfer transition may involve either one-or twoelectron transfer in the excited state, (40) so both the oneelectron potential, for the couple FL./FW (where known) and the two-electron potential, FL,/Fl,d, are listed. It should be noted that while the correlation is reasonably good for vm versus the two-electron potential for most of the flavins, the correlation is much better if the one-electron potential for 5-deaza-FAD is used. This is in keeping with the known thermodynamic instability of the 5-deazaflavin radical (39). The correlation observed is that expected for a chargetransfer interaction in which oxidized flavin serves as the charge-transfer acceptor (40). The donor is most probably the ionized persulfide species, CoA-S-S-, and precedent for a -S-. . . FAD charge-transfer complex comes from studies with lipoamide dehydrogenase and glutathione reductase. The long wavelength absorption generated on two-electron reduction of these flavoenzymes has been attributed to a chargetransfer interaction between a cysteinyl thiolate residue and oxidized FAD (41)(42)(43)(44)(45). This absorbance is abolished on protonation of the thiolate (44, 45).

Physiological and Metabolic
Significance-Butyryl-CoA dehydrogenases from most mammalian sources (4,28,46) and from M. elsdenii (3) are green. Glutaryl-CoA dehydrogenase, a similar flavoprotein from Pseudomonas fluorescens is also green (29). It is most likely that the greening ligand in each is a tightly bound CoA persulfide. The in vivo formation of such a ligand must involve some form of sulfur donor.
In the case of butyryl-CoA dehydrogenase in M. elsdenii, which is a strict anaerobe, the CoA persulfide could be produced by the enzyme itself according to the following scheme: In the course of the experiments reported here, it was observed that a large excess of sulfide can reduce both general acyl-CoA dehydrogenase and butyryl-CoA dehydrogenase as

TABLE I11
Dependence of charge-transfer transition on flavin oxiddonreduction potential Apoprotein was reconstituted with the analogs shown and spectra were recorded in 0.7 ml of 50 m~ KPi, pH 7.6 at 10 "C, before and after the addition of 10 d of a mixture of 20 mM CoASH and 20 mM N a 8 , which had been preincubated for 1 h at room temperature.  (15).
In the case of "an tissues, a more specific sulfur donor may be involved, since even low levels of sulfide are toxic (21). A well documented sulfur donor is rhodanese (e.g. Refs. 21,47,and 48) which is found in large quantities in liver and kidney (21). No physiological function has yet been found for this enzyme, but it is known to catalyze a variety of reactions in vitro (21), such as Rhodanese + 2 S-SOg-+ rhodanese"& + 2 S G -RhodanewS? + 2 CNe rhodanese + 2 CNS-Rhodanese is mainly associated with the mitochondria in vivo (21).
3-Mercaptopyruvate sulfurtransferase is present in rat liver and other tissues (21). The enzyme converts 3-mercaptopyruvate to pyruvate and elemental sulfur:

CN-CNS-
this reaction may involve the formation of an enzyme persulfide (49).
The CoA persulfide is tightly bound to butyryl-CoA dehydrogenase and, in vivo, it may therefore present a block to dehydrogenase activity unless a sufficient quantity of substrate is present to displace the greening ligand. It is therefore possible that the CoA persulfide could serve as a regulatory ligand for butyryl-CoA dehydrogenase; the enzyme would be "activated" (by displacement of the CoA persulfide, an "inhibitor") only when a sufficient, threshold quantity of substrate was present.
The discovery that a CoA persulfide is responsible for the 710 nm absorption band in native butyryl-CoA dehydrogenase provides a satisfying solution to a number of long standing puzzles. The possible physiological significance of such a tightly bound ligand is intriguing and needs to be fully explored. Equally intriguing are the molecular mechanisms by which the enzyme is "greened" in vivo and by which the normal unstable persulfide is stabilized when complexed with butyryl-CoA dehydrogenase. Acknowledgment-We would like to thank Nick Rhodes for Tunning the samples on the amino acid analyzer.