Spectroscopic characterization of the number and type of iron-sulfur clusters in NADH:ubiquinone oxidoreductase.

The number and type of iron-sulfur clusters present in the NADH dehydrogenase of the mammalian respiratory chain were studied by a combination of low temperature magnetic circular dichroism (MCD) and quantitative electron paramagnetic resonance spectroscopies. MCD was used with the high molecular weight, soluble enzyme, and EPR was used with both the purified enzyme and Complex I (NADH:ubiquinone oxidoreductase). The results of the EPR experiments of the two types of preparations agreed with each other, as well as with the data in the literature for various types of membrane-bound preparations. The two methods gave concordant results showing the presence of one binuclear and of three tetranuclear NADH-reducible iron-sulfur clusters. Earlier studies using the cluster extrusion technique indicated a higher ratio of binuclear to tetranuclear clusters which may be explained by cluster interconversion during the extrusion process.

The number and type of iron-sulfur clusters present in the NADH dehydrogenase of the mammalian respiratory chain were studied by a combination of low temperature magnetic circular dichroism (MCD) and quantitative electron paramagnetic resonance spectroscopies. MCD was used with the high molecular weight, soluble enzyme, and EPR was used with both the purified enzyme and Complex I (NADH:ubiquinone oxidoreductase). The results of the EPR experiments of the two types of preparations agreed with each other, as well as with the data in the literature for various types of membrane-bound preparations. The two methods gave concordant results showing the presence of one binuclear and of three tetranuclear NADHreducible iron-sulfur clusters. Earlier studies using the cluster extrusion technique indicated a higher ratio of binuclear to tetranuclear clusters which may be explained by cluster interconversion during the extrusion process.
The number, type and spatial distribution of iron-sulfur clusters in the NADH:ubiquinone oxidoreductase (Complex I) segment of the mitochondrial respiratory chain is still the subject of considerable uncertainty (for recent reviews see Refs. [1][2][3][4]. Information concerning the constituent iron-sulfur clusters has thus far been restricted to EPR studies of Complex I, both intact (5)(6)(7)(8) and after resolution into various subfractions and subunits (5,8-lo), and EPR and iron-sulfur core extrusion studies of soluble NADH dehydrogenase preparations (7,11,12). The results are consistent in finding only iron-sulfur clusters of the [2Fe-2SI2+*'+ and [4Fe-4S]2+J+ type but disagree as to the relative number of each type of cluster.
EPR studies of Complex I indicate the presence of one or two [ZFe-ZS] and three [4Fe-4S] clusters. Four clusters are reduced by the natural substrate, NADH, and these are labeled N-1 to N-4. N-1 is assigned to a [2Fe-2SI1+ center on the basis of the observed g values (g = 2.02, 1.94, and 1.92) and spin relaxation properties (observable >30 K). The reported EPR spin quantitations for this signal range from 0.4 (6,13) to 0.8 (5,8) spins/FMN. Ohnishi and co-workers (8,(14)(15)(16)  distinguish it from a similar EPR signal that is attributed to a non-NADH-reducible [2Fe-2S] cluster, N-la, which is only observed at lower redox potentials. While N-la is easily reduced by dithionite plus benzyl viologen in inner membrane preparations, lower potential mediators are reported to be required for reduction in Complex I (8). EPR signals from the remaining three clusters, N-2 to N-4, have been resolved by a combination of potentiometric titrations and computer simulations of spectral line shapes (6,17 (15,(17)(18)(19). This center exhibits extremely fast spin relaxation such that it is observable only at temperatures below 10 K and at high microwave powers. In view of the reported spin quantitations for this center, less than 0.06 spins/FMN (6,20) and less than 0.25 spins/FMN (17), it seems probable that this center is not an intrinsic component of NADH dehydrogenase.
In contrast, iron-sulfur core extrusion experiments using the high molecular weight soluble enzyme (12) indicate a composition with [2Fe-2S] and [4Fe-4S] clusters in an approximate 2:l ratio. Recent EPR studies on fragments of Complex I have also been interpreted to suggest a 2:l ratio (10). Three possible explanations for these apparently contradictory results need to be considered. First, the interpretation of the existing EPR data on Complex I and the soluble high molecular weight enzyme may be erroneous. Second, the intact enzyme may contain two or more EPR-silent [2Fe-2S] clusters as a result of inaccessibility to reductants, or redox potentials outside the normal range, or spin coupling between neighboring centers. Third, [4Fe-4S] clusters may break down to give [2Fe-2S] clusters under the conditions used for core extrusion and resolution of subfractions of Complex I.
To assess the validity of these possible explanations, we have taken a different approach to the complex problem of determining the iron-sulfur cluster composition of NADH dehydrogenase, i.e. that of low temperature magnetic circular dichroism (MCD') spectroscopy. This technique has recently proven effective in determining the iron-sulfur composition in a range of multicluster metalloenzymes, e.g. succinate de-hydrogenase (21), fumarate reductase (22), and nitrate reductase (23). In particular it is effective in identifying and monitoring paramagnetic iron-sulfur clusters that are EPR-silent as a result of zero-field splitting or weak intercluster spin interaction.
In this paper we report the results of low temperature MCD studies of high molecular weight soluble NADH dehydrogenase and quantitative E P R studies of this soluble enzyme preparation and Complex I. Complex I was not amenable to low temperature MCD investigation due to relatively high levels of cytochrome impurities which give rise to intense MCD transitions that obscure those from paramagnetic ironsulfur clusters. A comparative E P R study of Complex I and of the high molecular weight soluble enzyme used in this work was necessary because of the paucity of the available EPR data for this type of enzyme preparation. The close similarity in the EPR characteristics of the soluble enzyme with those of Complex I attest to the intactness of this type of preparation and enable the conclusions of the MCD studies to be extended to include Complex I.

MATERIALS AND METHODS
Complex I was prepared from beef heart mitochondria according to the published procedure (24). Samples for spectroscopic study were resuspended in anaerobic pH 8.0 Tris/sucrose/histidine buffer containing 0.66 M sucrose, 50 mM Tris chloride, and 1 mM histidine. High molecular weight soluble NADH dehydrogenase was isolated from the inner membrane of beef heart mitochondria as described by Paech et al. (25). Samples for spectroscopic study were in pH 7.8, 30 mM potassium phosphate buffer containing 50% (v/v) ethylene glycol, which facilitates the formation of optically clear glasses on freezing samples for MCD investigations. NADH dehydrogenase activity was assayed as described by Singer (26). The concentrations of FMN and FAD, and of non-heme iron were determined by the published procedures (27,28). Protein was determined by the biuret method (29) after precipitation with trichloroacetic acid. The enzyme concentrations used in quantifying MCD, EPR, and absorption spectra are based on the concentration of FMN. Both Complex I and soluble NADH dehydrogenase were shipped overnight in ice from San Francisco to Baton Rouge and samples for spectroscopic measurements were prepared immediately on arrival. Samples were kept on ice under an argon atmosphere at all times.
Sample handling involving addition of buffered solutions of dithionite and/or NADH was performed anaerobically under argon on a gas handling line. Photochemical reduction was accomplished by adding degassed stock solutions of EDTA and deazaflavin (100 and 1.0 mM, respectively, in Tris-HC1 buffer) to final concentrations of 10 mM and 20% of that of the enzyme, respectively, followed by illumination at 0 "C for 30 min, using a 200-W mercury-xenon arc lamp. This procedure was carried out in an EPR tube under argon and samples were frozen in liquid nitrogen immeditely after termination of illumination.
Low temperature MCD and EPR and room temperature absorption measurements were carried out using the instrumentation described in Ref. 21. Quantitations of EPR spectra were performed at 1-mW microwave power using either 1 mM CuEDTA (>30 K) or 1 m M metmyoglogin cyanide (<30 K) as the standard (30). Checks were made to ensure that the conditions used for spin quantitation did not produce power saturation in either the sample or standard.

RESULTS
Analytical and assay data for the samples of Complex I and high molecular weight soluble NADH dehydrogenase used in this work are presented in Table I. Based on the FMN content and the turnover number in the NADH-ferricyanide reductase assay (31,32), as compared with previous preparations, the Complex I sample is estimated to be approximately 60% pure.
However, the iron content, 16 g atoms of Fe/mol of FMN, which is at the lower end of the range reported for Complex I preparations (16-24 g atoms of Fe/mol of FMN (5,24,31,(33)(34)(35), and the observed absence of FAD indicates that the EPR signals indicative of other components of the respiratory chain were observed. EPR spin quantitations for each center and the maximal total spin quantitation for the complete spectrum are given in Table 11. Analogous spectra to those shown in Fig. 2, with almost identical spin quantitation data, were also observed for samples reduced with NADH without dithionite or via photochemical reduction in the presence of EDTA/deazaflavin.
As previously noted (6), the cluster observed above 30 K, N-1, is not reduced by dithionite but is readily reduced by NADH. The spin quantitation data indicate that this cluster is present at a concentration approximately stoichiometric with FMN, which is in agreement with the findings of Ohnishi et al. (8) and Orme-Johnson et al. ( 5 ) , but approximately double the values reported by Albracht's group (6,13). While the number of distinct species contributing to N-1 EPR spectrum has been the subject of vigorous debate, the current consensus is that is can be satisfactorily simulated as a single rhombic component (36). No increase in the intensity of the N-1 EPR signal was observed for samples photochemically reduced with or without NADH. We conclude that either a non-NADH reducible [2Fe-2S] cluster, N-la, is not present in these samples or that it cannot be reduced under these conditions. Below 30 K, three additional iron-sulfur EPR signals are observed, N-2, N-3, and N-4. They can be resolved into distinct rhombic components by differences in both power saturation characteristics and reducibility by dithionite. In the presence of NADH, all three clusters are found to have a spin quantitation approximately stoichiometric with FMN. N-2, which is known to exhibit the highest potential of any of the iron-sulfur clusters in Complex I, is the only cluster that is fully reduced by dithionite in the absence of NADH. The axial EPR spectrum of N-2, gll = 2.054, g, = 1.928, is most clearly seen for dithionite-reduced Complex I a t 20 K (see Fig. 1) since N-1 is not reduced and N-3 and N-4 are only partially reduced in this sample. In contrast to N-3 and N-4, the EPR signal of N-2 is strongly power-saturated at 8 K and 1-mW microwave power. Resolution of the two remaining rhombic EPR signals, N-3 (g = 2.103,1.93-1.94, and 1.885) and N-4 (g = 2.032, 1.92-1.93, and 1.862), was first accomplished by spectral simulations together with detailed studies of power saturation properties (6). The middle resonance of both clusters is obscured by overlap with components of the N-2 and/or N-1 signals, and the low field resonance of N-4 is only clearly observed when N-1 is oxidized, as in the dithionite-reduced sample (see Fig. 1). Accurate spin quantitation, based on a single "absorption-shaped" resonance, is only possible for N-3 because the low field signal is well separated from the other components. Hence, the value quoted in Table I1 for the spin quantitations of N-4 are probably overestimates because of overlap between the g = 1.885 (N-3) and g = 1.862 (N-4) Fig. 4 were also observed in samples reduced with NADH alone or via photochemical reduction. Clearly, all four of the iron- sulfur EPR signals observed in Complex I are also seen in the reduced soluble enzyme. While the g values and relaxation properties are essentially unchanged, the EPR signals, in particular that of N-2, are significantly broader, resulting in less well resolved spectra. Estimates of the spin quantitation of individual centers, see Table 11, are correspondingly less accurate. However, the spin quantitation data does attest to the same cluster stoichiometry as in Complex I. Compared to Complex I, dithionite alone did effect partial reduction of N-1 as well as N-3 and N-4 in the soluble enzyme. Similar EPR spectra to those shown in Figs. 3 and 4 have also been reported for samples of Complex I after treatment with cholate and salt which removes approximately 50% of the phospholipids (7). As for the soluble high molecular weight enzyme studied in this work, such samples do not exhibit ubiquinone reductase activity. Therefore, in agreement with EPR studies of other types of preparation, the results reported here indicate that removal of phospholipids and solubilization of NADH dehydrogenase leaves the EPR-detectable ironsulfur clusters essentially unchanged except for slight modifications in their molecular environment.
For reasons that are presently unclear, some samples of the soluble enzyme used in this work exhibited very broad, rhombic-type EPR signals at temperatures below 20 K on reduction  signals from the various subfractions resolved from Complex I by chaotropic agents (10). Spin quantitation data for the entire spectrum at 70 and 15 K, as well as the corresponding low temperature MCD spectra, were similar to those observed for samples exhibiting the EPR signals shown in Figs. 3 and 4. The conditions required to elicit these different types of EPR signal and the relationship between them will be the subject of future investigations.

MCD Studies of Soluble, High Molecular Weight NADH
Dehydrogenase- Fig. 5 shows the room temperature absorption and MCD spectra at 1.53, 4.22, 7.5, and 90 K and 4.5 T for dithionite-reduced soluble NADH dehydrogenase. The UV-visible absorption monotonically increases towards the UV except for bands centered around 558 and 426 nm which are characteristic of a low spin Fe(1I) cytochrome impurity. This diamagnetic cytochrome impurity is manifest in the low temperature MCD spectra as sharp temperature-independent bands between 500 and 560 nm and a very weak temperatureindependent, derivative-shaped Soret band centered around 416 nm. The multiplicity and position of the dominant temperature-independent bands suggests a mixture of cytochromes c and b as the likely contaminants (37, 38). The temperature-dependent derivative-shaped band between 400 and 455 nm is indicative of a trace of high spin (S = 2) Fe(I1) heme (39, 40), and this i s confirmed by magnetization data at 440 nm (data not shown). The remaining temperature-dependent bands in the MCD spectrum are attributed to paramagnetic iron-sulfur clusters. To determine the nature of the electronic ground state responsible for these transitions, magnetization curves were recorded in regions of minimum overlap with the cytochrome MCD bands. The magnetization data were not wavelength-dependent and a typical set of data, recorded at 750 nm, is shown in Fig. 6. Since the experimental points are fit to a good approximation by theoretical curves constructed for the average g value for the observed EPR signals, g,, = 1.96, MCD intensity at these wavelengths must "' O0I  Fig. 3. a, room  temperature absorption. b, MCD recorded at 1.53, 4.22, 7.5, and 90 K, magnetic field, 4.5 T, pathlength, 0.17 em. All transitions increase in intensity with decreasing temperature except for the temperatureindependent region around 550 nm. c, MCD difference spectrum; 1.53 K spectrum minus 90 K spectrum. clusters (41). Low temperature MCD spectra have now been recorded for a wide range of proteins that contain well characterized binuclear or tetranuclear iron-sulfur clusters: [2Fe-2SI1+ clusters in ferredoxins from spinach and Spirulina maxima (42,43), adrenodoxin (43), Rieske protein from Thermus thermophilus; red protein from Clostridiump~teuricmum,3 fumarate reductase from Escherichia coli (22), and mammalian succinate dehydrogenase (21); and [4Fe-4SI1+ clusters in ferredoxins from C. pasteurianum (44) and Desulfovibrw africanus (45), the nitrogenase Fe-proteins from Azotobacter uinelandii4 and C. paste~rianum,~ and mammalian electron transfer flavoprotein dehydrogenase.6 All [4Fe-4SI1+ clusters investigated thus far exhibit broad positive bands centered at 750 and 550 nm, multiple positive bands between 350 and 500 nm, and negative features around 625 and 320 nm. The MCD intensities of the most intense transitions are in the range 50 to 75 M" cm" for each [4Fe-4SI1+ cluster, for the conditions used in Fig. 5c. The dominant feature in the low temperature MCD spectrum of all [2Fe-2S11+ clusters investigated thus far is an intense negative band centered around 325 nm (At = 200 to 300 M-' cm" for the conditions used in Fig. 5c). To establish more clearly the form of the MCD from the paramagnetic iron-sulfur clusters and hence determine the relative contributions of binuclear and tetranuclear clusters to the MCD intensity, the temperature-independent cytochrome contributions were subtracted out by taking the difference between the 1.53 and 90 K spectra (Fig. 5c). Bearing in mind that the most intense band at around 440 nm in Fig. 5c originates from a heme impurity, we conclude that the intensity and form of the low temperature MCD spectrum of the paramagnetic ironsulfur clusters in dithionite-reduced NADH dehydrogenase is consistent with the presence of 2 or 3 [4Fe-4SJ1+ clusters and maximally 0.3 of a [2Fe-2SI1+ cluster. This result is entirely consistent with the quantitative EPR analysis discussed above and hence gives no indication of any EPR-silent, paramagnetic iron-sulfur clusters in this sample.
Room temperature UV-visible absorption and low temperature MCD spectra for soluble high molecular weight NADH dehydrogenase reduced by dithionite in the presence of NADH are shown in Fig. 7 with NADH alone gave identical EPR spectra, these samples were not amenable for study by MCD since their low temperature spectra were dominated by the intense temperaturedependent transitions from low spin Fe(II1) (5' = Yz) cytochrome impurities. In the presence of dithionite the cytochrome is reduced, giving identical bands to those observed in the sample reduced with dithionite alone. Taking the difference between the 1.53 and 90 K MCD spectra reveals the form of the temperature-dependent MCD transitions (Fig.   7c). A comparison of the temperature-dependent MCD transitions for dithionite-and dithionite plus NADH-reduced NADH dehydrogenase is shown in Fig. 8a. The additional intensity in spectral regions associated with paramagnetic iron-sulfur clusters indicates that addition of NADH results in further reduction of iron-sulfur clusters. MCD magnetization plots at 750 and 360 nm of the dithionite plus NADHreduced enzyme (data not shown) were analogous to that shown in Fig. 6 and are consistent with an S = Y Z ground state for all paramagnetic iron-sulfur clusters. The form of the low temperature MCD spectrum for the additional ironsulfur clusters that are reduced on addition of NADH is obtained by taking the difference between these spectra ( All transitions increase in intensity with decreasing temperature except for the temperature-independent region around 550 nm. c, MCD difference spectrum; 1.53 K spectrum minus 90 K spectrum.

DISCUSSION
Recent developments in iron-sulfur cluster chemistry and biochemistry have served to highlight the pitfalls of assigning the number and type of iron-sulfur clusters in multicluster enzymes on the basis of EPR and/or core extrusion data. Additional spectroscopic evidence on intact enzymes from techniques such as Mossbauer or low temperature MCD that are capable of detecting EPR-silent, paramagnetic clusters, is necessary before a secure assessment of the type of the constituent iron-sulfur clusters can be made. This is well illustrated by the controversy that existed concerning the number and type if the iron-sulfur clusters in mammalian succinate dehydrogenase and the eventual resolution of the dilemma (46).
The MCD results reported here for soluble high molecular weight NADH dehydrogenase confirm the cluster assignment based on the quantitative EPR studies and give no indication of any EPR-silent, paramagnetic iron-sulfur clusters. Both the MCD and EPR studies are best rationalized in terms of four NADH-reducible iron-sulfur clusters, each approximately stoichiometric with FMN: one binuclear, N-1; and three tetranuclear, N-2, N-3, and N-4. In view of the close similarity in both the form and quantitation of the EPR spectra of the soluble enzyme and Complex I, we conclude that the same clusters are also present in the particulate preparation. No EPR evidence for any additional iron-sulfur clusters was found for either type of preparation.
Unfortunately the usefulness of the Fe/FMN ratio in determining the stoichiometry of the multiple iron-sulfur clusters in both Complex I and soluble, high molecular weight NADH dehydrogenase is severely curtailed by the wide range of literature values: 16-24 Fe/FMN for Complex I (5, 24,31,[33][34][35]; and 17-22 for the high molecular weight, soluble enzyme (25,32). The lability of the FMN and/or iron-sulfur clusters combined with the variable purity of preparations from different laboratories is presumably responsible for this variation. Our own experience with Fe analysis of Complex I over a period of 20 years leads us to favor Fe/FMN ratios at the low end of the published range, and the sample used in the work reported here contained 16 Fe/FMN. We attribute the higher values for the purified enzyme (17.5 Fe/FMN for the original procedure (32) and 22 Fe/FMN for the rapid procedure (25)) to impurities or partial loss of FMN during purification. Since the NADH-reducible iron-sulfur clusters in Complex I only account for 14 Fe/FMN, the possibility of one or more non-NADH-reducible clusters needs to be considered. Ohnishi et al. (8) have reported evidence for one non-NADH-reducible [2Fe-2S] cluster in Complex I, N-la, with a midpoint potential around -500 mV. However, we were unable to elicit any increase or change in either the binuclear or tetranuclear iron-sulfur cluster EPR signal in samples of Complex I or soluble enzyme that were reduced photochemically using EDTA and deazaflavin. The deazaflavin radical is a powerful one-electron reductant with a midpoint potential around -650 mV at pH 7.0 (47). We conclude that cluster Nla is either inaccessible to reduction by the deazaflavin radical or absent from the samples used in this work. To further address this aspect, photochemical reduction experiments in the presence of a range of low potential mediator dyes are planned.
EPR studies of various resolved subfractions and subunits of Complex I were interpreted in terms of there being three tetranuclear and five or six binuclear iron-sulfur clusters in intact Complex I (22)(23)(24) Fe/FMN) (10). The MCD results reported here rule out the possibility that some [2Fe-2S] clusters are reducible by NADH but remain EPR-silent due to weak intercluster spin coupling. Thus, this cluster composition requires the existence of four or five [2Fe-2S] clusters that are not reduced by NADH as a result of inaccessibility to reductant or redox potentials lower than -650 mV. Such a situation seems to us to be extremely improbable. Moreover, we have serious reservations concerning any conclusions about the cluster composition of Complex I that are based on EPR studies of resolved subfractions. Our reasons are 2-fold. First, the EPR spectra of the resolved subfractions bear little resemblance to those of Complex I and indicate major perturbations of the iron-sulfur clusters on fractionation. Second, only a very small percentage of the total Fe in each subfraction can be accounted for by the iron-sulfur E P R signals. Hence, the observed E P R signals represent minor species which may, for the most part, be artifacts of the fractionation procedure.
The results of iron-sulfur cluster extrusion experiments on the soluble, high molecular weight NADH dehydrogenase (12) are also in conflict with the iron-sulfur cluster stoichiometry found by EPR and MCD spectroscopies. The extrusion studies were essentially quantitative with all the Fe being recovered in the form of two tetranuclear and four binuclear iron-sulfur clusters. The only rational explanation for this observation is that one [4Fe-4S] cluster fragments to produce two [2Fe-2S] clusters under the extrusion conditions. Precedent for such a cluster conversion, although not under extrusion conditions, comes from the recent studies of nitrogenase Fe-protein (48). Spectroscopic investigations designed to determine whether or not such a cluster conversion can occur in NADH dehydrogenase under conditions of protein unfolding are currently in progress in this laboratory.
In summary, we conclude that both MCD and EPR studies of NADH dehydrogenase concur in finding one binuclear and three tetranuclear NADH-reducible iron-sulfur clusters in both Complex I and soluble high molecular weight NADH dehydrogenase. While no positive evidence for additional iron-sulfur clusters was forthcoming in this work, previous EPR studies of Complex I suggest the presence of an additional non-NADH-reducible [2Fe-2S] cluster. A cluster composition for Complex I of two binuclear and three tetranuclear is consistent with the observed Fe/FMN ratio, 16 F e F M N , and the 1:1 stoichiometry of non-heme Fe to acid-labile sulfur.