Theaflavins inhibit the ATP synthase and the respiratory chain without increasing superoxide production

Abstract

Four dietary polyphenols, theaflavin (TF1), theaflavin-3-gallate (TF2A), theaflavin-3′-gallate (TF2B) and theaflavin-3,3′-digallate (TF3), have been isolated from black tea, and their effects on oxidative phosphorylation and superoxide production in a model system (E. coli) have been examined. The esterified theaflavins were all potent inhibitors of the membrane-bound ATP synthase, inhibiting at least 90% of the activity, with IC₅₀ values in the range of 10-20 μM. ATP-driven proton translocation was inhibited in a similar fashion, as was the purified F₁-ATPase, indicating that the primary site of inhibition was in the F₁ sector. Computer modeling studies supported this interpretation. All four theaflavins were also inhibitory towards the electron transport chain, whether through Complex I (NDH-1) or the alternative NADH dehydrogenase (NDH-2). Inhibition of NDH-1 by TF3 appeared to be competitive with respect to NADH, and this was supported by computer modeling studies. Rates of superoxide production during NADH oxidation by each dehydrogenase were measured. Superoxide production was completely eliminated in the presence of about 15 μM TF3, suggesting that inhibition of the respiratory chain by theaflavins does not contribute to superoxide production.

Introduction

Tea, a widely consumed beverage in the world, has long been considered to be beneficial to health (for reviews see [1-3]). Tea is distinguished by high levels of polyphenols, which are thought to be primarily responsible for any benefits to health. (For a comprehensive review see [4]). Green tea polyphenols consist primarily of catechins, mainly including (−) epicatechin (EC), (−) epicatechin gallate (ECG), (−) epigallocatechin (EGC), (−) epigallocatechin gallate (EGCG), (+) gallocatechin, and (+) catechin [5]. During the production of black tea, the catechins are enzymatically oxidized to produce theaflavins, which are orange or orange-red in color and possess a benzotropolone skeleton. The major theaflavins in black tea are theaflavin (TF1), theaflavin-3-gallate (TF2A), theaflavin-3′-gallate (TF2B), and theaflavin-3,3′-digallate (TF3). Theaflavins contribute importantly to properties of black tea including its color, ‘mouthfeel’ and extent of tea cream formation [6]. The oxidation process also forms larger, and less well characterized, polymers, referred to as thearubigins. The chemistry of these compounds has been reviewed recently [7]. The chemical structures of the theaflavins and EGCG are shown in Fig. 1.

Figure 1.

Chemical Structure of Polyphenols used in this Study. (A) (−) epigallocatechin gallate (EGCG) (B) Theaflavin (TF1) (C) Theaflavin-3-gallate (TF2A) (D) Theaflavin-3′-gallate (TF2B) (E) Theaflavin-3,3′-digallate (TF3)

The potential health benefits associated with tea consumption have been partially attributed to the antioxidative property of tea polyphenols [8]. Reaction mechanisms and products of reactions with peroxyl radicals [8, 9], hydrogen peroxide [10, 11] and superoxide [12, 13] have been investigated. The antioxidant potential of theaflavins in particular has been examined in vitro [14, 15]. In mammals, most superoxide is produced in the mitochondria, and principally from the electron transport chain [16]. Dioxygen can be reduced to superoxide at both Complex I and Complex III, and this preferentially occurs when key components of the electron transport chain are reduced. Many, but not all, inhibitors of the electron transport chain cause an increase in the rate of superoxide production [17]. Inhibitors of the ATP synthase can have the same effect, since when ATP synthesis is inhibited, the proton motive force builds up, and the electron transport chain slows down [18].

Several plant polyphenols, including tea catechins, have been shown to be inhibitors of the mitochondrial ATP synthase [19-21]. In the lab of Walker, the bovine F₁ sector of the ATP synthase was co-crystallized with three different polyphenols: resveratrol, piceatannol, and quercetin [22, 23]. All three were bound inside the enzyme along side the rotary shaft of the gamma subunit, and not at the nucleotide binding sites. The ATP synthase from E. coli, known to be a valuable model for the mitochondrial enzyme, was also found to be inhibited by many similar compounds. Treatment with piceatannol, morin, silymarin, baicalein, silibinin, rimantadin, amantidin, and epicatechin resulted in complete inhibition, while resveratrol, quercetin, quercetrin, quercetin-3-β-D-glucoside, hesperidin, chrysin, kaempferol, diosmin, apigenin, genistein, and rutin were partially inhibitory, in the range of 40–80% [24, 25].

In this paper we have extended previous work by showing that polyphenols from black tea, the theaflavins, can also inhibit the ATP synthase from E. coli, and that they are more potent than is EGCG. In similar fashion, these compounds are also inhibitory towards Complex I (NDH-1) of the electron transport chain. No evidence of increased production of superoxide was found in the presence of these inhibitors.

Materials and methods

Chemicals

EGCG, NADH, deamino-NADH, capsaicin, hexamine ruthenium chloride, and other chemicals were from Sigma-Aldrich (St. Louis, MO USA). ATP was from Roche (Indianapolis, IN USA). ACMA (9-amino 3-chloro 2-methoxy acridine) and MitoSOX were purchased from Invitrogen (Carlsbad, CA USA). Theaflavins were prepared as described previously [15]

Strains used

The wild type strain was 1100 [26]. BA14 is a derivative of strain 1100 with a deletion of the nuo genes, and so lacks all NDH-1 activity[27]. MWC215 contains a knockout of the ndh gene, and so lacks all NDH-2 activity [28]. It was generously provided by Dr. Bob Gennis (U. Illinois, Urbana IL USA). Strain DK8 with plasmid pFV2 [29] was used for purification of F₁.

Purification and preparations

1100 cells (wild type) or BA14 cells were grown in LB medium containing 1% tryptone, 0.5% yeast extract and 0.5 % NaCl at 37 °C. DK8/pFV2 cells were grown in LB medium supplemented with ampicillin at 0.05 g/L. MWC215 cells were grown in rich medium containing 3% tryptone, 1.5% yeast extract, 0.15% NaCl, and 1% (v/v) glycerol at 37 °C, supplemented with chloramphenicol at 0.05 g/L. Membrane vesicles for assays of ATP synthase and purified F₁ were isolated as described by Volkov, et al [30]. F₁-ATPase was purified essentially as described in Wise [31], except that the final purification of the ATPase was achieved by Ni-NTA chromatography. The enzyme was eluted at 180 mM imidazole. Membrane vesicles for assays of NDH-1 and NDH-2 were prepared at pH 6.0 as described previously [27].

Functional assays

The rate of ATP hydrolysis was measured by quantifying the release of inorganic phosphate using a colorimetric assay. After the equilibration of 200 μg of membrane vesicles from wild type strain 1100 or 40 μg of purified F₁ with test compounds in 500 μl buffer containing 5 mM MgCl₂, 50mM Tris (pH 7.5) for 5 min at room temperature, the assay was started by adding 5 mM ATP and 4 mM MgCl₂. The reaction was stopped after 3 min by addition of 250 μl of 10% SDS and 250 μl of color reagent [32]. The total phosphate was measured spectrophotometrically at 700 nm after 10 min. Phosphate at zero time point was determined by adding 10% SDS to the ATPase cocktail before adding the protein.

ATP-dependent ACMA fluorescence quenching was measured in 2 ml of buffer containing 50 mM MOPS and 10 mM MgCl₂ (pH 7.3). After the equilibration of 200 μg of membrane vesicles from wild type strain 1100 with tested compounds for 5 min, the reaction was initiated by addition of 0.1 mM of ATP and 1 μM of the dye under constant stirring at 37 °C, and stopped by addition of 3 μM FCCP . Fluorescence was excited at 410 nm and monitored at 490 nm using a Perkin-Elmer LS-5B luminescence spectrometer [29].

NADH oxidase activity in membrane vesicles was measured using a Beckman Coulter DU800 spectrophotometer by following NADH oxidation at 340 nm (ε=6.22 mM⁻¹ cm⁻¹) at 30 °C. For the NADH oxidase activity of NDH-1 from strain 1100 and NDH-2 from strain BA14, the reaction medium (1 ml) contained 50 mM MOPS (pH 7.3), 10 mM MgCl₂, and 200 μg of membrane vesicles. For the NADH- HAR (hexamine ruthenium) reductase activity, the reaction medium (1 ml) contained 50 mM MOPS (pH 7.3), 10 mM MgCl₂, 200 μg of membrane vesicles from strain MWC215, 10 mM KCN, and 1 mM HAR. The reaction was started by adding 75 μM deamino-NADH (for strain 1100) or 250 μM NADH (for strain BA14 or MWC215) after the equilibration of membrane vesicles with the test compounds for 5 min. The plotted concentrations represent the final concentration in the assay chamber.

Inhibition kinetics

The modes of inhibition and inhibition constants were determined by means of Dixon plots [33] and Cornish-Bowden plots [34], in which the reciprocal velocity, 1/v or s/v was plotted against the inhibitor concentration, i, at three values of s, the substrate concentration. Other details of the assays were indicated in the Figure legends.

Detection of superoxide with MitoSOX

Generation of superoxide by membrane vesicles from strain MWC215 or BA14 was estimated using the MitoSOX™ Red superoxide indicator in opaque-walled 96-well tissue culture plates. The reaction medium (0.2 ml) in each well contained 50 mM MOPS (pH 7.3), 10 mM MgCl₂, 200 μg of membrane vesicles. The reaction was started by adding 5 μM MitoSOX™ reagent and 1.25 mM NADH after the incubation of membrane vesicles with tested compounds for 5 min. SOD (150 units) was added in the control experiments. The plates protected from light were incubated for 15 min at 30 °C. The increase in fluorescence was monitored on a Molecular Devices Gemini XPS plate reader (Sunnyvale, CA) at 510/580 nm.

Under all assay conditions described above, without the addition of membrane vesicles, the five polyphenols tested had no relevant effects on the absorption or fluorescence of the indicators used.

Molecular docking

ChemDraw ultra 8.0 software [Chemical Structure Drawing Standard; Cambridge Soft corporation, USA (2003)] was used for construction of compounds which were converted to 3D structures using Chem3D ultra 8.0 software and the constructed 3D structures were energetically minimized by using VEGA 2.3.2 [35, 36]. Bovine F₁-ATPase (Protein Data Bank code: 1efr [37], resolution: 3.10 Å, co-crystallized with efrapeptin) and the hydrophilic domain of respiratory complex I from Thermus thermophilus (pdb code: 3i9v [38], resolution: 3.10 Å) were obtained from the Protein Data Bank (http://www.rcsb.org). The files were edited by using Chimera software [39]. All the water and inhibitors were removed. AutoDock 4.2 was employed for all docking calculations [40]. Polar hydrogen atoms and Gasteiger charges were added to the protein (receptor) model prior to calculation of grid maps using a grid box with a npts (number of points in xyz) of 62–62–62 Å box. For F₁-ATPase docking, the box spacing was 1.0 Å and grid center was designated at dimensions (x, y, z): 107.703, 74.005 and 73.815. For complex I docking, the box spacing was 0.7 Å and grid center was designated at dimensions (x, y, z): −29.238, 8.223 and 37.853. Ten runs were generated by using Lamarckian genetic algorithm searches. Default settings were used with a population size of 150, a maximum number of 2.5 × 10⁶ energy evaluations, and a maximum number of 2.7 × 10⁴ generations. A mutation rate of 0.02 and a crossover rate of 0.8 were chosen. The best ligand–receptor structure for each of the docking simulation was chosen based on the lowest energy.

Results

Inhibition of the ATP synthase

The four major theaflavin compounds: theaflavin (TF1), theaflavin-3-gallate (TF2A), theaflavin-3′-gallate (TF2B), and theaflavin-3, 3′-digallate (TF3) were tested as inhibitors of the membrane-bound F₁F_o-ATPase, along with (−) epigallocatechin gallate (EGCG). Several catechins, including EGCG have been previously shown to be inhibitors of the rat liver mitochondrial ATPase [20], but the theaflavins are somewhat larger than EGCG (see Fig. 1), and therefore might not be capable of binding to the same sites. Membrane vesicles were prepared from wild type cells, and ATP hydrolysis was measured in the presence of increasing amounts of the polyphenols. The results are summarized in Table 1. The theaflavins with gallate esters (TF2A, TF2B, TF3) all inhibited to an extent of 90-95 % with IC₅₀ values in the range 10-20 μM. TF1 and EGCG were somewhat less potent, with IC₅₀ values of about 60 and 30 μM respectively. Complete inhibition data are shown in Supplementary Fig. 1. Previous work has indicated that the sites of inhibition by polyphenols are in the F₁ sector of the ATP synthase [22]. Since the theaflavins are twice as large as many other polyphenols, it was possible that they might bind to other regions of the enzyme. For that reason, isolated F₁-ATPase was examined as well. These parallel results are also presented in Table 1 and Supplementary Fig. 1. In each case, a very similar pattern of results was obtained, except that the IC₅₀ values were about 10-fold lower using the isolated F₁-ATPase.

Table 1.

Inhibition of ATPase Activity by Theaflavins

	Membrane-bound ATPase Activitya		F1-ATPase Activityb
Inhibitor	IC50	Maximum inhibition	IC50	Maximum inhibition
	(μM)c	(%)c	(μM)c	(%)c
TF1	60	85	4.0	85
TF2A	20	95	3.0	90
TF2B	15	95	1.5	95
TF3	10	90	0.7	90
EGCG	30	95	4.5	90

^aMembrane-bound ATPase activity was about 0.7 μmoles ATP/min/mg protein

^bF₁-ATPase activity was about 1.8 μmoles ATP/min/mg

^cIC₅₀ concentrations and maximum inhibition % were determined from two replicates of 5-6 concentrations of inhibitor. The data are displayed in Supplementary Figure1.

Proton translocation requires the coupling of ATP hydrolysis to the proton translocation apparatus of the membrane-bound F_o sector. This can be detected by monitoring the fluorescence quenching of an acridine dye, ACMA. In this assay, the rate and extent of fluorescence quenching is indicative of the generation of a proton gradient across the membrane. Each of the four theaflavins was tested as an inhibitor of ATP-driven proton translocation, using membrane vesicles, and the results are presented in Fig. 2. In each case, the inhibition of proton translocation seemed to follow that of the inhibition of ATP hydrolysis (Table 1), and so there was no evidence for additional inhibitory binding sites in the membrane sector of the enzyme. The binding site for several polyphenols in the F₁-ATPase has been well-established by the co-crystallization of the bovine enzyme with resveratrol, piceatannol, and quercetin, each binding to the same position inside the enzyme near the membrane-distal end [22]. Considering the size of the theaflavin polyphenols, especially the digallate ester (TF3), it seemed unlikely that they could bind to the same site as the smaller flavonol quercetin. Computer modeling and docking studies were carried out with each of the theaflavins, along with EGCG, to see where they might bind in the F₁-ATPase. Using the bovine F₁-ATPase structure (Protein Data Bank file 1efr) [37], stripped of ligands as the template, quercetin was docked to a position very close to the actual binding site discovered by co-crystallization studies (see Fig. 3A). In contrast, TF3 was docked to a totally different site, closer to the membrane, and overlapping with the actual binding site of efrapeptin [37] (see Fig. 3B).

Figure 2.

Inhibition of ATP-driven Proton Translocation by Polyphenols. Proton translocation by membrane vesicles is indicated by the quenching of the fluorescence of ACMA upon the addition of ATP. Membranes without the addition of polyphenols are indicated by CK. In each case, the fluorescence quenching could be eliminated by the addition of FCCP (carbonyl cyanide p-(trifluoromethoxy)phenylhydrazone), a protonophore (results not shown). (A) TF1, (B) TF2A, (C) TF2B, (D) TF3. The final concentrations of inhibitors are indicated in the panels. The results shown are typical from 2 assays of 2 different preparations of membranes.

Figure 3.

Comparison of Actual and Predicted Inhibitor Binding Sites in Bovine F₁-ATPase. F₁ is shown in gray wireframe, the gamma subunit is in thick wireframe, and the inhibitors are shown in black spacefilling. (A) The model shown is from Protein Data Bank file 2jj2. (B) The results from a docking study of TF3 and the F₁-ATPase from Protein Data Bank file 1efr are shown. In both cases the inhibitors are bound to the gamma subunit, but the TF3 is docked much lower, and below the nucleotide binding sites. The TF3 site in panel B is accessible from the bottom of the structure, as shown, rather than from the top, as in panel A.

Inhibition of Electron Transport

Since the polyphenol resveratrol has previously been shown to inhibit the mitochondrial respiratory chain [41, 42], the theaflavins were also tested as inhibitors of the electron transport chain in membrane vesicles. E. coli has two distinct NADH dehydrogenases in its plasma membrane [28, 43]. Both use the oxidation of NADH to reduce ubiquinone, which is re-oxidized by one of two quinol oxidases, bd and bo [44]. NDH-1 is homologous to the mitochondrial Complex I. It is a multi-subunit, proton translocating enzyme, and is similar to the ATP synthase, in that it has two large sectors, one embedded in the membrane, and another peripheral to it. The peripheral sector, binds NADH, while the membrane sector is involved with ion translocation. In contrast, NDH-2 is a single polypeptide that is monotopically bound to the membrane and does not translocate ions. In E. coli, they can be distinguished by the use of strains in which one enzyme is knocked out genetically, or by the use of deamino-NADH, which only NDH-1 is able to use. Each NADH dehydrogenase was tested with the four theaflavins and EGCG, and the results of the inhibition are presented in Table 2. Complete inhibition data are shown in Supplementary Fig. 2. For NDH-1 the IC₅₀ values for the 5 polyphenols range from about 10-50 μM, and the inhibition reaches 90-95%. In contrast, the theaflavins inhibit NDH-2 with lower IC₅₀ values in the range of 1-4 μM, while the extent of inhibition is 75-95%. However, the IC₅₀ for EGCG is about the same with respect to both NADH dehydrogenases. Clearly, the inhibitors have different targets in these two respiratory chains.

Table 2.

Inhibition of NADH Oxidase Activity by Theaflavins

	NDH-1 Deamino-NADH oxidasea		NDH-2 NADH oxidaseb
Inhibitor	IC50	Maximum inhibition	IC50	Maximum inhibition
	(μM)c	(%)c	(μM)c	(%)c
TF1	45	95	3.0	75
TF2A	15	95	3.5	95
TF2B	8	95	2.5	90
TF3	7	90	1.5	95
EGCG	20	90	20	65

^aNDH-1 activity using deamino-NADH was about 0.5 μmoles deamino-NADH/min/mg protein

^bNDH-2 activity using NADH was about 0.25 μmoles NADH/min/mg protein

^cIC₅₀ concentrations and maximum inhibition % were determined from two replicates of 5-6 concentrations of inhibitor. The data are displayed in Supplementary Figure2.

Next, the modes of inhibition by TF3 were examined, by varying the amounts of NADH and TF3, and these results are presented in Fig. 4. Both Dixon [33] plots (1/v vs. s) and Cornish-Bowden [34] plots (s/v vs. s) are shown, and the assays were done with both a wild type strain, using deamino-NADH, and with a mutant strain lacking NDH-1. The results are consistent with competitive inhibition, since the lines are parallel in the Cornish-Bowden plot, and intersect in the upper left quadrant in the Dixon plot. The K_i can be estimated to be about 0.4 μM from the wild type strain, and about 1 μM from the mutant strain. In contrast, NDH-2 shows uncompetitive behavior under inhibition by TF3, as shown in Fig. 5. The Dixon plots are parallel lines, and the Cornish-Bowden lines intersect in the upper left quadrant. From these plots the K_i′ can be estimated to be 0.8 μM.

Figure 4.

Inhibition of NDH-1 by TF3 is Competitive with Respect to NADH. NADH oxidase activity was measured with deamino-NADH using wild type membranes (1100) (panels A, C), and with NADH using strain MWC215, which lacks NDH-2 activity (panels B, D). Assays were carried out with 3 different concentrations of deamino-NADH or NADH, with increasing amounts of TF3 added. The results are plotted according to Cornish-Bowden (panels A, B) or Dixon (panels C, D). The units of s/v are μM TF3 (μmoles NADH/min/mg protein) ⁻¹, and the units of 1/v are (μmoles NADH/min/mg protein) ⁻¹. Each point shown represents the mean of two measurements.

Figure 5.

Inhibition of NDH-2 by TF3 is Uncompetitive with Respect to NADH. NADH-oxidase activity was measured with NADH using strain BA14, which lacks NDH-1 activity. Assays were carried out with 3 different concentrations of NADH, with increasing amounts of TF3 added. The results are plotted according to Cornish-Bowden (panels A) or Dixon (panels B). The units of s/v are μM TF3 (μmoles NADH/min/mg protein) ⁻¹, and the units of 1/v are (μmoles NADH/min/mg protein) ⁻¹. Each point shown represents the mean of two measurements.

To support the finding that the NDH-1 was competitively inhibited by TF3, computer modeling studies were carried out using the peripheral sector of the Thermus thermophilus Complex I (Protein Data Bank file 3i9v) [38]. Results showed that the highest affinity binding occurred on the surface of the enzyme at a site that partially overlapped with the NADH binding site. In addition, the activity of NDH-1 was assayed by the use of an artificial electron acceptor, hexamine ruthenium (HAR), which is thought to draw off electrons before they reach the membrane sector. In this case, using TF3 as the inhibitor, the results showed uncompetitive inhibition (results not shown), where the Dixon plots have parallel lines, and the Cornish-Bowden lines intersect in the upper left quadrant. The K_i′ was estimated to be 0.16 μM. This appeared to be in conflict with the earlier finding that TF3 inhibits competitively at the NADH site.

Modulation of the Rate of Superoxide Production

Inhibitors of electron transport typically cause an increase in superoxide production during the consumption of NADH. Since TF3 was found to be a rather potent inhibitor of both NDH-1 and NDH-2, rates of superoxide production were measured in membrane vesicles with increasing levels of TF3. The results, presented in Fig. 6 showed that NADH oxidation by NDH-1 caused higher rates of superoxide production than by NDH-2. Addition of superoxide dismutase reduced the fluorescence, and so indicated the amount that was due to superoxide. When TF3 was added at levels in the range of the IC₅₀ value, superoxide did not increase, but rather decreased. This shows that while TF3 is an inhibitor of the electron transport chain in E. coli membranes, it does not cause an increase in superoxide production.

Figure 6.

The Effect of Increasing Levels of TF3 on Superoxide Production by NDH-1 and NDH-2. Superoxide was measured by the increase in fluorescence of the Mitosox reagent, in the absence of NADH (CK) and in the presence of superoxide dismutase (SOD), which reflect the background fluorescence. To differentiate between NDH-1 and NDH-2, strains MWC215 was used for NDH-1 (shaded bars) and strain BA14 was used for NDH-2 (open bars). The results shown are the mean of three measurements with the error bars indicating the standard deviation.

Discussion

Flavonoid polyphenols are dietary compounds from plants that are known to interact directly with enzymes. The ATP synthase was previously found to be inhibited by several simple polyphenols, including resveratrol, quercetin, and EGCG [19, 20, 23-25]. In this report theaflavins, a class of polyphenols found exclusively in black tea, were also shown to be inhibitors of the ATP synthase. The four theaflavins tested varied in size according to the presence of 0, 1, or 2 gallate esters. The apparent affinities, as indicated by IC₅₀ values, had a four-fold range of values, and correlated well with overall size. The importance of the gallate group is also suggested by the inhibitory properties of EGCG. It is smaller than TF1, but contains both a gallate ester and a galloyl group, and has an IC₅₀ value that is similar to that of TF1. Furthermore, in another inhibition study of the E. coli ATP synthase [25], epicatechin, lacking both the galloyl and gallate groups, was found to have a substantially larger IC₅₀ of 4 mM. Similar to the results here, rat liver mitochondrial ATP synthase was found to be inhibited by EGCG with an IC₅₀ of 17 μM [20].

The extents of inhibition also correlated with the size of the polyphenols. That is probably related to the fact that the ATP synthase is a rotary motor, and many of its inhibitors appear to be mechanical inhibitors. For example, the polyphenols that have been-co-crystallized with the enzyme, such as quercetin or resveratrol are found to bind between the shaft of the rotary gamma subunit and the stationary alpha and beta subunits, which bind the nucleotides [22]. Since the theaflavins are somewhat larger than quercetin, and much larger than resveratrol, it could be expected that they would not bind to the same site, even though they are related chemically. Indeed, the computer modeling studies suggested a different binding site, one which overlaps with a previously identified site for the antibiotic efrapeptin, from crystallographic studies. This site is accessible from the opposite end of the enzyme, and it is a region that undergoes large conformational changes during rotary catalysis. It is possible that even with full occupancy, some of the smaller polyphenols fail to inhibit completely because they do not totally block rotation, but only hinder it.

Assays carried out with the isolated F₁-ATPase demonstrated that the site of inhibition was located in the F₁ sector of the enzyme, rather than F_o. The lower IC₅₀ values obtained relative to the membrane-bound enzyme might reflect complexities in the activity of the enzyme. The rate of ATP hydrolysis by the membrane-bound enzyme is quite sensitive to pH [45] and other conditions [46, 47] that cause it to be released from the membrane sector. The results shown do not preclude additional bindings sites in the membrane sector. Other inhibitors, such as the antibiotic oligomycin, are known to act there. However, the assays of proton translocation showed no evidence for additional inhibitor sites in the membrane sector. While it cannot be ruled out that lower affinity sites exist, the polarity of the theaflavins, especially those with gallate esters, might be too great to allow binding in the lipid phase.

Two different enzymes allow NADH to initiate electron transport in E. coli, and both were found to be inhibited by all 5 polyphenols tested. In the case of NDH-1, the IC₅₀ values followed the same pattern as with the ATP synthase: affinity correlated with size, except that TF1 had a larger IC₅₀ than did EGCG. This again indicates the likely importance of the gallate groups in binding. The computer docking of TF3 suggested important interactions of the benzotropolone group and one of the gallates (3′) with the enzyme, supporting that conclusion and the relatively low IC₅₀ for TF2B. Overall, the polyphenols had relatively high extents of inhibition (90-95%), which is consistent with a competitive mode of inhibition, as was found for TF3, in which substrate binding is precluded. The inhibition of NDH-2 followed the pattern of larger theaflavins having lower IC₅₀ values, but in this case, EGCG had the highest IC₅₀ value of all by a factor of 5-10. This suggests importance of the benzotropolone group, in conjunction with at least one gallate. Overall, the extents of inhibition were somewhat lower than for NDH-1, with values in the range of 65-95%. Since the mode of inhibition by TF3 was determined to be uncompetitive, it suggests that for some of the inhibitors, the enzyme-inhibitor-NADH complexes retained some activity.

It was not expected that TF3 would be a competitive inhibitor with respect to NADH, since it does not resemble NADH structurally. On the other hand, it also does not closely resemble ubiquinone, so there was no basis to predict that it would bind in the membrane sector. The finding of its competitive mode of inhibition of NADH oxidase activity was in conflict with the finding that it was uncompetitive with respect to NADH using the HAR reductase assay. The former result requires TF3 binding in the absence of NADH, while the later requires binding in the presence of NADH. An important consideration is that the mechanism of the HAR reaction is unknown. A recent paper by Birrell [48] suggests that HAR might draw off electrons through the protein from one of the Fe-S centers that is close enough to the surface. Given such uncertainties about this artificial electron acceptor, we favor the interpretation that the inhibition is competitive with respect to NADH. Supporting that view are the findings by computer modeling that a potential binding site exists that is partially overlapping with the entrance to the NADH binding site. In this way the binding of TF3 could block the binding of NADH, without significant structural similarity to it.

Many inhibitors of the electron transport chain are significant contributors to superoxide production in the mitochondrion, because they tend to keep particular electron carriers in a reduced state. For example, some Complex I inhibitors are known to increase the rate of superoxide production [17, 49]. Superoxide has been suggested to be formed at various sites in Complex I, including at the flavin, at an Fe-S center, or at a ubisemiquinone site [50-54]. The mode of the inhibition by TF3 is consistent with its lack of stimulation of superoxide production. A competitive inhibitor will tend to prevent reduction of the electron carriers, rather than to keep them in a reduced state. In addition, theaflavins are known to be superoxide scavengers [14]. In this study, TF3 caused a decrease in superoxide production while it inhibited the electron transport chain in E. coli membranes at levels in the range of the IC₅₀ value.

The bioavailability of catechin compounds in the human body is thought to be low [55, 56], notwithstanding the possibility of higher bioavailability for some catechins such as EGCG [57], or in some tissues such as mouth and esophagus [58]. Thus, given the micromolar IC₅₀ values measured here, and previous studies on the health benefits of theaflavins as antioxidants in vivo, we hypothesized that theaflavins had dual effects on cells in a dose-dependent way. At low doses, theaflavins play an important role as antioxidants, and the physiological consequences of the inhibition of ATP synthase or Complex I by theaflavins might be limited. At high dosage, theaflavins might have some effect on the electron transport chain. These results suggest the possibility of theaflavins used as drugs targeting Complex I or ATP synthase of some pathogens such as Mycobacterium tuberculosis, or in some diseases such as obesity [59]. In addition, the results presented here suggest that there are modes of inhibition of these enzymes that might be common to a variety of polyphenols, or related compounds, found in the human diet.

List of Abbreviations:

ACMA

9-amino 3-chloro 2-methoxy acridine

EC

(-) epicatechin

ECG

(-) epicatechin gallate

EGC

(-) epigallocatechin

EGCG

(-) epigallocatechin gallate

HAR

hexamine ruthenium

IC₅₀

concentration of inhibitor required to achieve 50% of the maximal level of inhibition

NDH-1

NADH dehydrogenase 1, or Complex I

NDH-2

alternative NADH dehydrogenase

TF1

theaflavin

TF2A

theaflavin-3-gallate

TF2B

theaflavin-3′-gallate

TF3

theaflavin-3,3′-digallate