Reconstitution of Escherichia coli thioredoxin reductase with 1-deazaFAD. Evidence for 1-deazaFAD C-4a adduct formation linked to the ionization of an active site base.

The flavin prosthetic group (FAD) of thioredoxin reductase has been replaced by 1-deazaFAD (carbon substituted for nitrogen at position 1). Reduction of 1-deazaFAD-thioredoxin reductase by four electrons proceeds in two stages having midpoint potentials that are separated by 0.063 V. Two-electron reduced 1-deazaFAD-thioredoxin reductase (EH2) has spectral characteristics that are different from both the fully oxidized and fully reduced enzyme. The fluorescence of the 2-electron reduced enzyme shows a mixture of two EH2 species. The spectrum of one EH2 species has a single absorption peak (lambda max, 414 nm; epsilon 414, 8750 M-1 cm-1) which is similar to the spectrum of 1-deazaFAD-C-4a adducts (referred to as the 414-nm absorbing species). In the other EH2 species the electrons are in the dithiol, and it has an oxidized 1-deazaFAD spectrum (referred to as the 550-nm EH2 species). The equilibrium between the two EH2 species of 1-deazaFAD-thioredoxin reductase is pH dependent, forming more of the 414-nm absorbing species as the pH is lowered. The pH dependence suggests the presence of an active center base having a pK of 7.41 on the 414-nm EH2 species and a thiol of pK 6.73 on the 550-nm EH2 species. These pK values are similar to the pK values determined for native enzyme having a disulfide or a dithiol (7.59 and 6.98, respectively). Thus, the pH dependence of the equilibrium between the two EH2 species of 1-deazaFAD-thioredoxin reductase is further evidence for an active site base with an ionization behavior that is linked to the chemical state of the active site disulfide moiety. The nature of the linked ionization is consistent with a thiol base ion pair formed upon disulfide reduction.

disulfide Thelander, 1968;Moore et al., 1964;Ronchi and Williams, 1972). It is thought that the electrons flow from NADPH to the FAD, from the FAD to the disulfide, and from the dithiol to the disulfide of thioredoxin. Thioredoxin reductase is analogous to lipoamide dehydrogenase and glutathione reductase in having an FAD and active center disulfide and in transferring electrons between pyridine nucleotide and disulfide substrates Holmgren, 1980). The active center disulfide sequences of glutathione reductase from yeast (Jones and Williams, 1975) and erythrocyte (Krohne-Ehrich et al., 1977) are highly homologous to the disulfide sequences of lipoamide dehydrogenase from E. coli (Burleigh and Williams, 1972) and pig heart (Matthews et al., 1974;Perham 1972,1974). However, the active center disulfide sequence of E. coli thioredoxin reductase (Ronchi and Williams, 1972;Thelander, 1970) shows no homology to the disulfide sequences of lipoamide dehydrogenase and glutathione reductase. A further difference between lipoamide dehydrogenase and glutathione reductase and thioredoxin reductase is in the separation of the oxidation-reduction midpoint potentials (E,) of the FAD and disulfide couples. The FAD couple of both lipoamide dehydrogenase and glutathione reductase has a much lower E,,, value than the disulfide couple (AE,, approximately 0.066 V from pH 5.5 to 7.6 in pig heart lipoamide dehydrogenase, Matthews and Williams, 1976) leading to a two-stage reduction; first the disulfide is reduced and then the FAD is reduced. The 2-electron reduced forms of lipoamide dehydrogenase and glutathione reductase are spectrally distinct enzyme species having a 530-nm absorbance band interpreted as a charge transfer complex between oxidized FAD as the acceptor and a thiol anion donor (Kosower, 1966;Searls and Sanadi, 1961;Massey and Ghisla, 1974). Studies from this laboratory suggest that the thiol anion charge transfer donor is stabilized in an ion-pair interaction with an essential base at the EH2* level in lipoamide dehydrogenase (Matthews and Williams, 1976;Matthews et al., 1977) and in glutathione reductase (Arscott et al., 1981).
The E,,, values of the FAD and disulfide couples of thioredoxin reductase are approximately equal Williams 1982, 1983). Thus, reduction of thioredoxin reductase appears to be a single stage process showing only a gradual bleaching of the FAD absorbance (isosbestic wavelength, 347 nm) throughout a 4-electron titration . NO charge transfer band is detected during the reduction of thioredoxin reductase even though studies demonstrate a thiol anion at the active site Williams 1982, 1983).
The present paper is a study of a derivative of thioredoxin reductase in which the FAD is replaced by 1-deazaFAD, an analogue of FAD having a more negative E,,, (Walsh et al., 1978). The flavin replacement effects a separation of E , between the flavin and disulfide couples which is similar to the separation determined for lipoamide dehydrogenase.

MATERIALS AND METHODS
Thioredoxin reductase was purified by a modification of the method of Pigiet and Conley (1977) using 1.0 M NaCl in place of NADPH to elute the enzyme from the 2',5'-ADP affinity column since aerobic turnover of NADPH leads to peroxide-mediated protein modification? NAD+ and NADH were purchased from Sigma, sodium dithionite was from Tridom Chemical Corp., and activated charcoal was from Atlas Chemical Industries. 1-DeazaFAD at the riboflavin level was a generous gift of the synthetic chemistry group (Ashton et al., 1977) of Merck, Sharpe and Dohme Research Laboratories to Dr. Christopher Walsh of Massachusetts Institute of Technology. The adenine dinucleotide, 1-deazaFAD, was a gift from Dr. Vincent Massey (University of Michigan) and was prepared using the partially purified flavokinase and FAD synthetase from Breuibacteriurn amrnoniagenes following the general procedures of Spencer et al. (1976).
Absorbance measurements were performed in a Cary 118C recording spectrophotometer interfaced to a PDP/8E computer (Williams et al., 1979) at 12 "C. Fluorescence measurements were performed in a Perkin-Elmer model 44B recording ratio spectrofluorimeter at 12 "C and were not corrected for monochrometer artifacts.
The FAD was resolved from thioredoxin reductase by adding 5.17 ml of 8.0 M guanidinium chloride, 0.1 M K2HPO-NaH2P04, 0.3 mM EDTA, pH 7.6, to 3.1 ml of 42.1 PM thioredoxin reductase in 0.1 M KZHP04-NaH2P04, 0.3 mM EDTA, pH 7.6 (Buffer A) at 5 "C. The visible absorbance spectrum indicated resolution of the FAD from the enzyme within the time of mixing. The FAD was removed from solution by direct addition of defined activated charcoal in 0.1 ml of Buffer A. Additions of charcoal were repeated (four additions) until no detectable FAD absorbance remained. The supernatant was dialyzed against Buffer A to remove the guanidinium chloride. The concentration of apothioredoxin reductase was quantitated by amino acid analysis.
Apothioredoxin reductase was reconstituted with FAD by the addition of excess FAD to apothioredoxin reductase in Buffer A followed by extensive dialysis to remove unbound FAD. The concentration of reconstituted enzyme was quantitated by the absorbance at 455 nm using an extinction coefficient of 11,300 M" Cm" . Apothioredoxin reductase was reconstituted with 1-deazaFAD by adding aliquots of 1-deazaFAD and monitoring the appearance of fluorescence (excitation, 552 nm; emission, 637 nm). When the fluorescence ceased to increase upon further additions of 1-deazaFAD, the reconstituted enzyme was dialyzed against Buffer A to remove unbound 1-deazaFAD. The extinction coefficient of 1-deazaFADthioredoxin reductase at the visible wavelength maximum (552 nm) is approximately 6800 M" cm" determined by dithionite titration. The native and reconstituted enzymes were assayed using the 5,5'dithiobis-(2-nitrobenzoic acid) coupled assay (Moore et aZ., 1964). Calculation of the Semiquinone Spectrum-The semiquinone spectrum was calculated from spectra obtained during an anaerobic sodium dithionite titration of 23 +M l-deazaFAD-thioredoxin reductase at pH 7.6 (spectra not shown but the following argument can be followed by reference to Fig. 1). The sepctrum prior to the final addition of dithionite contained the EH2 and EHI forms of enzyme as well as a substantial amount of the 1-deazaFAD semiquinone form as judged from the absorbance at 700 nm. This spectrum will be referred to as Spectrum A. The spectrum of semiquinone was calculated by correcting Spectrum A for the absorbance contributions of the EH, and EH, species. Spectrum A contained 54% EH, based on the fluorescence change upon the final addition of dithionite (excitation, 380 nm; emission, 490 nm). Spectrum A also contained 10.1% of the fluorescence of the oxidized 1-deazaFAD in EH2 (excitation, 552 nm; emission, 660 nm). The final addition of dithionite resulted M. E. O'Donnell and C. H. Williams, Jr., unpublished data.
in a complete loss of the absorbance at 700 nm, a loss of the EH, fluorescence, an increase in the E H 4 fluorescence, and produced an absorption spectrum of fully reduced 1-deazaFAD-thioredoxin reductase, EH4. Thus, two changes occur upon the last addition of dithionite: l) reduction of EH, to EH4 and 2) reduction of semiquinone to Although EH2 is an equilibrium mixture of at least two enzyme forms (i.e. oxidized 1-deazaFAD-dithiol and a C-4a thiol-1-deazaFAD adduct) it is a spectrally distinct mixture at a given pH. The absorbance contribution of EH, to Spectrum A was calculated as follows. TWO other spectra in the second phase of reduction (1.25 and 1.5 mol of dithionite/mol of 1-deazaFAD) had the same absorbance at 700 nm and presumably the same concentration of semiquinone. Thus, their difference spectrum, referred to as Spectrum B, yields the absorbance changes for EHz to EH4. The fluorescence of EH, decreased to 46% of the value before this dithionite addition. Based on the loss of EHz fluorescence, the reduction of EH, to EH4 in the final addition of dithionite (Spectrum A to EH,) resulted in absorbance changes equivalent to 22% of Spectrum B. Thus, 22% of Spectrum B was subtracted from Spectrum A to give a difference spectrum referred to as Spectrum C. Spectrum C is the calculated spectrum for a mixture of EH4 and semiquinone. The contribution of EH, to Spectrum C is the amount of E H 4 in Spectrum A (54%) plus the 10% of EH4 introduced by the subtraction of 22% of Spectrum B from Spectrum A (0.22 X 0.46 = 0.1). Thus, 64% of the absorbance of reduced 1-deazaFAD-thioredoxin reductase was subtracted from Spectrum B, and the resultant spectrum was normalized for enzyme concentration to yield the spectrum of semiquinone shown in Fig. 2. Since the reduction state of the disulfide/dithiol couple in the 1-deazaFAD-semiquinone cannot be determined it is assumed that the reduction state of the disulfide/dithiol couple does not significantly affect the semiquinone spectrum.
Correction of Spectra for Semiquinone at pH 7.6-The absorbance spectra were corrected for semiquinone by calculating the concentration of semiquinone from the absorbance at 700 nm using the extinction coefficient of 1774 M" cm" and subtracting the absorbance spectrum of the appropriate concentration of semiquinone using the calculated spectrum of semiquinone at pH 7.6 ( Fig. 2). The resulting spectrum is scaled for the concentration of enzyme as semiquinone.
Calculation of the Spectrum of EH-During the first phase of an anaerobic sodium dithionite titration at pH 7.6, the amount of absorbance at 700 nm was identical for two spectra that were separated by 0.57 eq of reduction. Thus, the concentration of semiquinone is identical at the two levels of reduction, and the difference spectrum gives the absorbance changes for 57% E to 57% EH,. The difference spectrum was extrapolated to 100% absorbance change for the first stage of reduction, EHz -E, and the spectrum of E (starting spectrum) was added to the extrapolated difference spectrum to yield the absorbance spectrum of E H z at pH 7.6 (Williams et al., 1979).
Estimation of the Spectrum of the 414-nm EHZ Species-The spectrum of the 414-nm EH, species was estimated as follows. The absorbance changes for the conversion of the 550-nm EH, species to the 414-nm EH, species were the difference between Spectrum 9 and Spectrum 2 in the experiment of Fig. 7 after correcting each for an 11% contribution of semiquinone. The extent of the shift of the 550nm EH, to the 414-nm EH2 was calculated to be 28% assuming an extinction coefficient at 550 nm of the 550-nm EH, and the 414-nm EH2 to be 6800 M" cm" and 600 M" cm", respectively. Since native thioredoxin reductase which has electrons trapped on the disulfide via complexation with phenylmercuric acetate (O' Donnell and Williams, 1983) has an extinction coefficient of the FAD close to that of native enzyme, the value of 6800 M" cm" was assumed for the 550nm EH, species. An extinction coefficient of 600 M" cm" for the 414-nm EH, species at 550 nm was assumed on the basis of the slight absorbance at 550 nm for the C-4a 1-deazaFAD adduct of the reaction between a-hydroxybutanoate and lactate oxidase (Entsch et al., 1980). This is consistent with the studies presented here which show that the 414-nm EH, species has little or no absorbance at 550 nm (see Fig. 5 and text). Using these extinction coefficients, the spectrum of EH, at pH 7.6 (see above) is calculated to consist of 80.7% of the 550-nm EH, species and 19.3% of the 414-nm EHz species. Thus, the spectrum of 100% 414-nm EHz species was obtained by adding the appropriate amount of the difference spectrum (Spectrum 2-Spectrum 9, Fig. 7) to the spectrum of EH, at pH 7.6. The spectrum of the 414-nm EH, species calculated by this method is shown as the dotted spectrum in Fig. 7. The use of different extinction coefficients at 550 nm for the two EHz species changes the extinction coefficient EH4.
by guest on July 9, 2020 http://www.jbc.org/ Downloaded from at 414 nm, but does not alter its general form.
Measurements of Oxidation-Reduction Midpoint Potentials-The E , values for the two phases of reduction of 1-deazaFAD-thioredoxin reductase were calculated as described in the Miniprint of O 'Donnell and Williams (1983). Briefly, the E, of the first phase, Ez, was calculated from absorbance measurements during an anaerobic NADH titration at pH 7.6, 12 "C. The equilibrium concentrations of E and EHz were calculated from the absorbance at 552 nm using extinction coefficients of 6800 M" cm" and 5500 M" cm" (from the calculated EHz spectrum), respectively. The equilibrium concentration of NADH was calculated from the residual absorbance at 340 nm after subtracting the absorbance contributions at 340 nm of E and EH2 using their extinction coefficients of 3100 and 3400 cm", respectively. The equilibrium concentration of NAD+ was the difference between the amount of titrant added and the equilibrium concentration of NADH. Semiquinone concentration was calculated from the absorbance at 700 nm using an extinction coefficient of 1775 M" cm". The absorbance measurements at 340 and 552 nm were corrected for semiquinone using the extinction coefficients of 3000 M" cm" and 300 M" cm", respectively. The E, of the first stage of reduction, Et, was calculated using the Nernst relationship: Ez =
The E,,, of the second phase, E,, was determined from a titration of fully reduced 1-deazaFAD-thioredoxin reductase (sodium dithionite) with NAD+. The NAD+ titrant solution was standardized as described in O' Donnell and Williams (1983). The equilibrium concentrations of E H z and EH4 were calculated from the absorbance at 552 nm using an extinction coefficient at 552 nm for EH4 of 174 M" cm". The equilibrium concentration of NADH was calculated from the residual absorbance at 340 nm ( € 3~~ 6220 M" cm") after subtracting the contributions of EHz and EH, using an extinction coefficient for EH, at 340 nm of 3550 M" cm". The equilibrium concentration of NAD+ was the difference between total NAD+ added and the equilibrium concentration of NADH. The absorbance measurements were corrected for semiquinone as in the calculations of E,.

RESULTS
Preparation of 1 -DeazaFAD-Thioredoxin Reductase-The FAD has been resolved from thioredoxin reductase using 5.0 M guanidine HC1 and removed from solution by direct addition of activated charcoal. An amino acid analysis of the dialyzed apoprotein showed a 30% loss of total protein presumably due to adsorption of apoprotein to the activated charcoal. The dialyzed apoprotein contained no detectable FAD absorbance and was 0.7% active with respect to native enzyme. Apothioredoxin reductase was 87% reconstitutable with FAD. The absorbance spectrum of the FAD-reconstituted enzyme was indistinguishable from that of native enzyme, and the reconstituted enzyme had 99% of the activity of native enzyme.
The binding of 1-deazaFAD to apothioredoxin reductase is accompanied by vibronic resolution, an approximately 20-nm shift of the wavelength maximum of the visible band (552 nm), and an increased extinction coefficient of the ultraviolet band relative to the visible band. These spectral changes are the same in nature as those of FAD bound to thioredoxin reductase. Reconstitution of apothioredoxin reductase with 1-deazaFAD was 90% complete based on an extinction coefficient at 552 nm of 6800 M" cm" (dithionite titration). 1-DeazaFAD-thioredoxin reductase had 6.8% the activity of native enzyme.
Reduction of 1 -DeazaFAD-Thioredoxin Reductase Occurs in Two Stages-An anaerobic sodium dithionite titration of 1-deazaFAD-thioredoxin reductase at pH 6.0 is shown in Fig.  1, A and B. 1-DeazaFAD-thioredoxin reductase is fully reduced by reaction with 2 eq of sodium dithionite (4 electrons). The reduction appears to occur in two stages producing an intermediate 2-electron reduced spectrum that has characteristics which differ from either oxidized or fully reduced en-zyme. The first stage of reduction shows a decrease in absorbance a t 552 nm, an increase at 414 nm, and an isobestic wavelength of 457 nm (Fig. 1A). The second stage of reduction shows a further decrease in absorbance at 552 nm including a decrease above 400 nm, a region which showed increasing absorbance during the first stage. The first three spectra of the second stage are isosbestic at 397 nm. A plot of the absorbance changes at 440 nm, a wavelength between the isosbestic wavelengths of the individual stages shows about 1 eq of reduction in each phase (inset to Fig. 1B).
The reduction of 1-deazaFAD-thioredoxin reductase is accompanied by a long wavelength-absorbing species. A long wavelength-absorbing species was observed by Spencer et al. (1977a) in reactions of reduced 1-deazariboflavin with oxygen and was attributed to the neutral form of the 1-electron reduced species of 1-deazariboflavin. The spectrum of the long wavelength species was calculated from spectra during a sodium dithionite titration at pH 7.6 (see under "Materials and Methods") and is shown in Fig. 2. The spectrum of the long wavelength species closely resembles the spectrum of the neutral radical of 1-deazaFAD-flavodoxin (Entsch et al., 1980). Thus, we assign the 700-nm species as being the neutral semiquinone of 1-deazaFAD-thioredoxin reductase. This assignment is consistent with the substantial quantity of neutral semiquinone that is formed during dithionite reduction of native thioredoxin reductase, especially at pH 6.0 (O' Donnell and Williams, 1983).
The lack of exact isosbestic points in reduction is probably due to the variable levels of semiquinone that are formed throughout the titration. This is suggested by Spectra 3-5 of the first stage of reduction which have similar absorbance at 700 nm and are isosbestic (Fig. LA) and by Spectra 10 and 11 of Fig. 1B which show the largest change in absorbance at 700 nm and corresponding large changes in the isosbestic wavelength. Spectra of a dithionite titration of 1-deazaFADthioredoxin reductase at pH 7.6 (similar in profile and 700nm absorbance to those at pH 6.0) were corrected for the semiquinone by subtraction of the calculated semiquinone spectrum at pH 7.6 (see under "Materials and Methods") and are shown in Fig. 3. The corrected spectra are isosbestic (first stage, 475 nm; second stage, 426 nm), further evidence for the reduction occurring in two discrete stages and for variable levels of semiquinone produced throughout the titration.
The two-stage reduction of 1-deazaFAD-thioredoxin reductase is analogous to the reduction of lipoamide dehydrogenase and glutathione dehydrogenase except that the charge transfer band at 530 nm, maximal in the first stage of reduction in lipoamide dehydrogenase and glutathione dehydrogenase, is replaced by an absorbance band at 414 nm which is maximal after the first stage of 1-deazaFAD-thioredoxin reductase reduction. Thus, a spectrally distinct intermediate species of 2-electron reduced 1-deazaFAD-thioredoxin reductase is formed during reduction and will be referred to as EH, by analogy to the intermediate charge transfer species of lipoamide dehydrogenase and glutathione dehydrogenase. The spectrum of 1-deazaFAD in fully reduced thioredoxin reductase (Spectrum 11, Fig. 1B; Spectrum 8, Fig. 3) has an absorbance maximum at 375 nm 5800 M" cm"). The spectrum of the neutral dihydro-1-deazaFAD in free solution has an absorbance maximum at 480 nm 2000 M" cm"), and the anion has no distinct peak above 330 nm (Spencer et al., 1977a). Thus, thioredoxin reductase must stabilize a different structural form of reduced 1-deazaFADH,. The several relevant tautomers and resonance forms of dihydro-l-deazaflavins are discussed in Spencer et al. (1977a). The pH independence of the reduced spectrum (pH 7.6 to pH 6.0; Figs FIG. 1. Spectra observed during an anaerobic dithionite titration of 1-deazaFAD-thioredoxin reductase at pH 6.0. A shows the absorption spectra recorded in the first stage of an anaerobic dithionite titration of 18.9 N M 1-deazaFAD-thioredoxin reductase in 1.2 ml of 0.1 M NaH2P04-K2HP04, 0.3 mM EDTA, pH 6.0,12 "C. 1, oxidized enzyme; 2, 0.6 eq; 3, 0.8 eq; 4, 1.0 eq; 5, 1.2 eq; 6, 1.4 eq of dithionite/l-deazaFAD. B shows the absorbance changes observed in the second phase of the same dithionite titration. 7, 1.6 eq; 8, 1.8 eq; 9,2.0 eq; 10, 2.2 eq; 11,2.4 eq of dithionite/l-deazaFAD. Inset, relationship of the extinction coefficient at 440 nm to equivalents of dithionite added. Residual oxygen accounts for 0.5 mol of dithionite/mol of 1-deazaFAD.  . 3. Spectra observed during a dithionite titration of 1-deazaFAD-thioredoxin reductase at pH 7.6 corrected for absorbance due to the 700-nm species. Absorptinn spectra recorded after additions of dithionite to 23.6 p~ I-deazaFAD-thioredoxin reductase in 1.2 ml of 0.1 M NaH2POI-K2HPO4, 0.3 mM EDTA, pH 7.6,12 "C. The spectra were corrected for the absorbance of semiquinone species as detailed under "Materials and Methods." Methyl viologen (0.5 PM) was added to facilitate reduction. The solid spectra indicate reduction up to 1 mol of dithionite/mol of 1-deazaFAD. The dashed spectra indicate reduction by more than 1 mol of dithionite/ mol of 1-deazaFAD. From top to bottom, oxidized enzyme, 0.30 eq, 0.62 eq, 0.81 eq, 1.25 eq, 1.50 e q , 1.86 eq, 2.0 eq of dithionitejmol of 1-deazaFAD. titration with NAD+. The value of E , was -0.362 V. The difference between El and E2 of 0.063 V predicts a n equilibrium constant of 170 for comproportionation. In lipoamide dehydrogenase, the values of Ez and El were -0.316 and -0.382 V, respectively (pH 7.6, 25 T), giving a difference of 0.066 V (Matthews and Williams, 1976). Thus, the E , values for the couples of native lipoamide dehydrogenase are about 0.020 V more negative than those of 1-deazaFAD-thloredoxin reductase, but the difference, E2 -El, is very similar for the two enzymes.
The E,,, value of 1-deazaFAD at pH 7.0, 25 "C, in free solution is 0.061 V more negative than the E,,, of free FAD (Walsh et al., 1978). El, the midpoint potential of 1-deazaFAD bound to thioredoxin reductase, is approximately 0.063 V more negative than the E, of the FAD in the dithiol form of thioredoxin reductase (O' Donnell and Williams, 1983

WAVELENGTH (nm) WAVELENGTH (nm)
FIG. 4. Fluorescence excitation spectra observed during an anaerobic dithionite titration of 1-deazaFAD-thioredoxin reductase at pH 6.0, 12 "C. Excitation spectra were recorded with emission at 660 nm. The fluorescence spectra were collected during the dithionite titration of Fig. 1. A , fluorescence excitation spectra for reduction up to 1 mol of dithionite/mol of 1-deazaFAD. B, fluorescence excitation spectra for reduction by more than 1 mol of dithionite/mol of 1-deazaFAD. the E, values of both FAD and 1-deazaFAD are lowered to approximately the same extent upon binding to apothioredoxin reductase indicating that any interactions between the protein and the N-1 position of the FAD do not significantly affect the E,,, of the flavin.
Nature of the EH2 Fluorescence; Evidence for Two Species-In contrast to the lack of fluorescence of 1-deazaFAD in free solution and bound to enzymes (Spencer et al., 1977133, both the oxidized and fully reduced forms of 1-deazaFAD-thioredoxin reductase are fluorescent. The oxidized 1-deazaFADthioredoxin reductase has 0.05% the fluorescence of FMN (pH 6.0; excitation maximum, 552 nm; emission maximum, 635 nm). The reduced form of 1-deazaFAD-thioredoxin reductase has 3.3% the fluorescence of FMN (excitation maximum, 392 nm; emission maximum, 575 nm), 66 times the fluorescence of oxidized 1-deazaFAD-thioredoxin reductase. The fluorescence excitation spectra of 1-deazaFAD-thioredoxin reductase, shown in Fig. 4, were recorded during the dithionite titration shown in Fig. 1. The quantum yields of the oxidized and reduced forms of 1-deazaFAD-thioredoxin reductase are approximately equal at an emission wavelength of 660 nm. Therefore, the excitation spectra obtained monitoring at this wavelength will be related to the absorbance spectra both in shape and in magnitude. Two features of the fluorescence results are important to note. At 414 nm, where EH, has more absorbance than E, the fluorescence excitation spectra show decreasing intensity throughout the first stage of the titration. This shows that the absorbance at 414 nm is caused by a nonfluorescent 2-electron reduced form of 1-deazaFAD-thioredoxin reductase. After reduction by 1 eq of dithionite, the 2-electron reduced enzyme is still quite fluorescent having an excitation spectrum very similar to that of E except for a blue shift of approximately 4 nm. Thus, the fluorophores at the E and EHz levels are similar. These observations indicate that there are at least two spectrally distinct species at the EH2 level, a fluorescent species having a maximum at approximately 550 nm, which will be referred to as the 550-nm EHz species, and a nonfluorescent species having an absorption maximum at 414 nm which will be referred to as the 414-nm EH, species. The relationship between the 552-nm fluorescence and the absorbance at 552 nm is shown in Fig. 5 for the experiment of Figs, 1 and 4  species have appreciable absorbance at 552 nm.

Effect of Phenylmercuric Acetate on EH2; the EH2 Species
Are in Rapid Equilibrium-The organic mercurial, phenylmercuric acetate, binds tightly to the active center dithiol of reduced thioredoxin reductase (O' Donnell and Williams, 1983). The addition of phenylmercuric acetate to 1-deazaFAD-thioredoxin reductase in the first stage of reduction caused a rapid loss of the 414-nm absorbance and a concomitant rise in the 552-nm absorbance in the time of mixing, about 15 s (Fig. 6). This shows that the 414-nm species is an EH2 species and that the 550-nm EH2 species and the 414-nm EH2 species are in rapid equilibrium. The 700-nm absorbance is essentially unchanged by phenylmercuric acetate addition, consistent with semiquinone as the 700-nm absorbing species.
Native thioredoxin reductase also has two 2-electron reduced enzyme species (O' Donnell and Williams, 1983). These species are enzyme forms in which the electrons reside on either the disulfide or the FAD and will be designated here as FADH,/disulfide and FAD/dithiol species. The FAD/dithiol EH, species is fluorescent' while the FADH,/disulfide EH, species is not fluorescent. The intramolecular equilibrium between the FADHZ/disulfide and FAD/dithiol species is rapidly shifted upon addition of phenylmercuric acetate toward the FAD/dithiol-phenylmercuric acetate complex.

-DeazaFAD-Thioredoxin Reductase
The fluorescence excitation spectrum of the 550-nm EH2 species shows that it is a 1-deazaFAD/dithiol species analogous to the FAD/dithiol species of native enzyme. However, several lines of evidence argue against the 414-nm EH, species as a 1-deazaFADH2/disulfide species. In native enzyme, the spectra of oxidized and reduced FAD are not altered by the oxidation-reduction state of the disulfide, whereas in 1-deazaFAD-enzyme, the absorbance and fluorescence properties of the 414-nm EH, species are very different from the spectral properties of fully oxidized enzyme or fully reduced enzyme, i.e. the spectrum of the 414-nm EH, relative to EH4 is red shifted 40 nm (see below) and has no detectable fluorescence. The E, values of the FAD and disulfide couples of fully oxidized native enzyme are approximately equal at pH 7.6 (-0.278 V and -0.282 V, respectively) (O' Donnell and Williams, 1983). Substantial amounts of both 1-deazaFAD-EH, species exist at the 2-electron reduced level. Thus, if the 414-nm EH, species contains 1-deazaFADH,, the E, of bound 1-deazaFAD would not be very different from the E,,, of the disulfide. This is not consistent with the low potential of 1-deazaFAD bound to thioredoxin reductase (see above).
The Equilibrium between the Two EH, Species Is pH-dependent-Lipoamide dehydrogenase from both pig heart and E. coli have a fluorescent EH, species and a nonfluorescent charge transfer EH2 species (Wilkinson and Williams, 1979).3 The electrons in these two spectrally distinct species of EHz in lipoamide dehydrogenase reside on the dithiol. The difference in the EH, species lies in their state of protonation. The nonfluorescent charge transfer EH, species has a thiol anion that forms a charge transfer complex with the FAD resulting in a new absorbance band at 530 nm (Kosower, 1966;Massey and Ghisla, 1974;Matthews and Williams, 1976;Wilkinson and Williams, 1979). The fluorescent EH, species contains a protonated thiol and is similar to oxidized enzyme in its absorbance and fluorescence properties. A 530-nm EH2 species is also observed in glutathione reductase (Arscott et  ~~ 1981). The extinction coefficient at 530 nm of the charge transfer EH2 species of lipoamide dehydrogenase and glutathione reductase is decreased as the pH is lowered, due to protonation of the thiol anion donor of the charge transfer complex (Matthews and Williams, 1976;Arscott et al., 1981). If the 414-nm EH2 species of 1-deazaFAD-thioredoxin reductase is a charge transfer species analogous to lipoamide dehydrogenase and glutathione reductase, the 414-nm absorbance of partially reduced 1-deazaFAD-thioredoxin reductase should decrease as the pH is lowered.
The pH dependence of the 414-nm absorbance of partially reduced 1-deazaFAD-thioredoxin reductase was measured by reducing the enzyme with slightly greater than 1 eq of dithionite in 0.010 M K2HP04-NaH2P04, 0.3 mM EDTA, pH 8.22, and the pH was lowered by 2341 additions of 0.5 M acetic acid. The pH of the anaerobic enzyme solution was determined in an aerobic control experiment by measuring the pH of the same volume of buffer titrated with 0.5 M acetic acid. The results are shown in Fig. 7. As the pH was lowered, the 414nm absorbance increased, and the 550-nm absorbance decreased. The observed spectral changes are opposite to those expected for a charge transfer EH2 species. Thus, the pH dependence of the 414-nm absorbance shows that the 414-nm EH, species is not a thiolate-to-1-deazaFAD charge transfer species.
The pH dependence of the equilibrium between the 550nm EH, species and 414-nm EH, species of 1-deazaFADthioredoxin reductase allows the spectrum of the 414-nm EHz species to be calculated (see under "Materials and Methods") and is shown as the dashed spectrum in Fig. 7. The characteristic twin maxima of oxidized 1-deazaFAD is replaced by a single peak in the 414-nm EH, species ( X , , , ,

nm; e4149
Model studies suggest that the transfer of electrons between flavin and disulfide proceeds via the covalent addition of a thiol at the C-4a position of the flavin. Thus, it was of interest to compare the calculated spectrum of the 414 nm EHz species with a known C-4a adduct of 1-deazaFAD. The reduced enzyme-substrate complex of 1-deazaFAD-p-hydroxyben-8750 M" Cm"). zoate hydroxylase reacts rapidly with oxygen to form a structure proposed to be a C-4a-1-deazaFAD hydroperoxide (Entsch et al., 1980) and also has a spectrum (Xmax, 395 nm; t395, 9500 M" cm") similar to that of the 414-nm EHz species.
Furthermore, the product of the reaction between 1-deazaFAD-lactate oxidase and 1-hydroxy-3-butynoate is proposed to be a C-4a/N-5 bridged adduct (Entsch et al., 1980) and results in a spectrum ( X , , , , 382 nm; 8300 M-l cm") which is similar to that of the 414-nm EH, species.
The spectrum of the 414-nm E H z species is compatible with formation of a thiol C-4a adduct of 1-deazaFAD exhibiting spectral characteristics similar to analogous adducts involving substitution of carbon or oxygen moieties at that position. All such adducts lead to a tetrahedral carbon atom at the C-4a position. The flavin-binding pocket in thioredoxin reductase is quite apolar as judged by the vibronic resolution of the absorbance spectrum of both FAD and 1-deazaFAD bound to the enzyme. Thus, the 19-to 32-nm red shifted position of the absorbance maximum of the putative 1-deazaFAD C-4a adduct in thioredoxin reductase relative to that for the 1-deazaFAD C-4a adducts of p-hydroxybenzoate hydroxylase and lactate oxidase could be the result of solvent effects.
A thiol-to-flavin C-4a adduct has also been observed in a derivative of lipoamide dehydrogenase wherein one of the active site thiols is alkylated by iodoacetamide (Thorpe and Williams, 1976a). This enzyme derivative is induced to form a flavin C-4a adduct upon binding NAD' (Thorpe and Williams, 1976b;Thorpe and Williams, 1981). Since NAD+ induces the FAD C-4a adduct in iodoacetamide-modified lipoamide dehydrogenase, NADP+ was added in a pH titration of partially reduced 1-deazaFAD-thioredoxin reductase analogous to the experiment of Fig. 7. The results were essentially the same as those of Fig. 7 indicating no significant induction effect of NADP+ on formation of the putative 1-deazaFAD C-4a adduct.
The respective concentrations of the 550-nm EH, species and 414-nm EH, species in the experiment of Fig. 7 were calculated from the 550-nm absorbance assuming extinction coefficients at 552 nm for the 550-nm EH, species and 414nm EHz species of 6800 M" cm" and 600 M" cm", respectively, and correcting for semiquinone assuming extinction coefficients of 300 M" and 1774 M" cm" a t 552 and 700 nm, respectively. The equilibrium constant, 550-nm EH2/414-nm EH,, and the equilibrium constant in the opposite direction are plotted as a function of pH in Fig. 8. The derivation and interpretation of the plot of Fig. 8 is described in O' Donnell and Williams (1983). The theoretical fits to the data yield pK values of 7.41 and 6.73 which are the values for ionizations on the 414-nm EH, species and 550-nm EH, species, respectively. The values for the internal equilibrium between the 414-nm EHp species and the 550-nm EH, species at the basic and acidic limbs of the theoretical fits to the data are 0.167 and 0.80, respectively. Thus, the 414-nm EHz species has an ionization with a higher pK than that of a group on the 550nm EH, species leading to a shift in equilibrium toward the 414-nm EH, species as the pH is lowered.

DISCUSSION
The reduction of 1-deazaFAD-thioredoxin reductase occurs in two stages that are separated in E,,, by 0.063 V. Two electrons react with the enzyme in each stage. The two-stage nature of the reduction is predicted from the low potential of 1-deazaFAD relative to FAD in solution (A&, 0.063 V). Thus, in native enzyme, the FAD and disulfide couples have approximately equal E,,, values and are reduced together whereas in 1-deazaFAD-thioredoxin reductase the disulfide is reduced (constituting the first stage, Ez = -0.299 V, pH 7.6, 12 "C) and then the 1-deazaFAD is reduced (second stage, E, = Lipoamide dehydrogenase and glutathione reductase show a two-stage reduction and a 2-electron reduced species having increased absorbance at 530 nm relative to both the oxidized and fully reduced enzymes. The 530-nm absorbance band is interpreted to be a thiolate-to-FAD charge transfer species. Two-electron reduced 1-deazaFAD-thioredoxin reductase, EH,, has increased absorbance a t 414 nm relative to both the oxidized and fully reduced enzyme. The 414-nm absorbance of the EH, species of 1-deazaFAD-thioredoxin reductase is not caused by a thiolate-to-1-deazaFAD charge transfer interaction since protonation of 1-deazaFAD-thioredoxin reductase EHz leads to increased amounts of the 414-nm absorbance. The fluorescence excitation spectra of 1-deazaFAD-thioredoxin reductase in the course of a dithionite titration show that EHz is a mixture of at least two spectrally distinct species, a fluorescent EHZ species that absorbs maximally at 550 nm and a nonfluorescent EH, species that absorbs maximally at 414 nm. The 414-nm EH2 species has a spectrum with a single absorbance peak ( X,,,, 414 nm; c414,8750 M" cm") suggestive of a 1-deazaFAD C-4a adduct. Since the putative C-4a adduct is formed upon reduction of the disulfide, the presumed C-4a substituent is one of the active site thiols. Since FAD and 1-deazaFAD have similar chemical properties (Spencer et al., 1977a) these results suggest an intermediate C-4a species in the transfer of electrons from FADH, to the disulfide in native enzyme as shown in Scheme 1. The C-4a of reduced flavin -0.362 V, pH 7.6, 12 "C). adds to the disulfide to form a thiol-to-flavin C-4a adduct and a free thiol. The electron transfer is completed concurrent with the deprotonation of the flavin N-5 to form oxidized flavin and a dithiol. An intermediate C-4a adduct species in the transfer of electrons between the FAD and thiols is consistent with model studies (Hemmerich, 1968;Hamilton, 1971;Gascoigne and Radda, 1967;Loechler and Hollocher, 1975;Yokoe and Bruice, 1975). A thiol-to-flavin C-4a adduct has also been observed in a derivative of lipoamide dehydrogenase wherein one of the active site thiols is alkylated by iodoacetamide (Thorpe and Williams, 1976a). This enzyme derivative is induced to form a C-4a adduct upon binding NAD+ (Thorpe and Williams, 1976b;Thorpe and Williams, 1981).
Measurements of the proton stoichiometry of reduction of the disulfide in native thioredoxin reductase indicate a base at the active center with an ionization behavior that is linked to the oxidation-reduction state of the disulfide (O' Donnell and Williams, 1983). The proton stoichiometry results were best fit by a model wherein enzyme with an oxidized disulfide had a group with a pK of approximately 7.59 and enzyme containing a dithiol had a group with a pK of approximately 6.98. The pH dependence of the equilibrium between the two EH2 species in 1-deazaFAD-thioredoxin reductase shows that the 414-nm EH2 species (which does not have a dithiol) has an ionization with a pK of approximately 7.41, and an ionization of a group on the 550-nm EH2 species (having a dithiol) has a pK of about 6.73. Thus, the pH dependence of the equilibrium between the EH2 species is further evidence for the ionization of an active site base linked to the chemical state of the disulfide moiety.
Studies on lipoamide dehydrogenase suggest that a base on oxidized enzyme has a pK value below 5.5 and is shifted up to 7.8 upon reduction of the disulfide. The shift in the pK of the base is explained by the formation of a thiol-base ion pair upon disulfide reduction. The linkage of the pK of a base to the chemical state of the disulfide (oxidized, reduced, or putative C-4a) in both 1-deazaFAD-thioredoxin reductase and native enzyme is consistent with a thiol-base ion pair as shown in Scheme 2 . We have indicated the pK 7.41 in Scheme 2 to be the nascent thiol which is in an ion pair interaction with the base, since the ion pair is favored in the apolar active site.
An ion pair is an attractive hypothesis because the ionization behavior of both the thiol and base of the ion pair fulfill needed functions predicted by the chemistry of thioredoxin reductase catalysis. Specifically, a thiolate is known to initiate 5 5 0 n m EH, 1 K,:0.800 K w 0 . 1 6 7 thiol-disulfide interchange reactions ( i e . transfer of electrons between the dithiol of thioredoxin reductase to the disulfide of thioredoxin). In addition, formation of the mixed disulfide between thioredoxin reductase and thioredoxin would be concerted with an increased acidity of the protonated base which could function as a proton donor to the free thiol of thioredoxin. Thus, both a nucleophilic thiol anion and a protonated base are required for efficient catalysis and are encompassed in the thiol-base ion pair model.
The results presented here on the mechanism of thioredoxin reductase suggest further similarities to the mechanisms of lipoamide dehydrogenase and glutathione reductase. Studies on iodoacetamide-modified lipoamide dehydrogenase suggest that native lipoamide dehydrogenase transfers electrons between the FAD and dithiol via a sulfur-to-flavin C-4a species. Characterization of 1-deazaFAD-thioredoxin reductase suggests a thiol-to-flavin intermediate in native thioredoxin reductase. Lipoamide dehydrogenase has a base whose pK is linked to the oxidation-reduction state of the disulfide and is the partner to a thiolate in an ion-pair interaction. The proton stoichiometry in the reduction of the disulfide of native thioredoxin reductase and the pH dependence of the equilibrium between the EH2 species of 1-deazaPAD-thioredoxin reductase disclose the presence of a base in thioredoxin reductase with an ionization behavior that is linked to the oxidationreduction state of the disulfide. The linked ionization is consistent with a thiol-base ion pair at the active site of thioredoxin reductase.
In lipoamide dehydrogenase, the ion pair thiol has a pK of about 4.8; and studies suggest that the base has a pK of less than 5.5 on oxidized enzyme (Matthews et al., 1977). This may indicate a thiol anion in thioredoxin reductase that has a greater intrinsic nucleophilicity than the thiol in lipoamide dehydrogenase. Thioredoxin reductase differs from lipoamide dehydrogenase and glutathione reductase in that the putative thiolate in thioredoxin reductase does not charge transfer to the FAD. This may be due to an incorrect juxtaposition of the thiolate relative to the FAD or a suboptimal ionization potential of the thiolate.