Exposing the Complex III Qo semiquinone radical

Abstract

Complex III Qo site semiquinone has been assigned pivotal roles in productive energy-conversion and destructive superoxide generation. After a 30 year search, a genetic heme b_H knockout arrests this transient semiquinone EPR radical, revealing the natural engineering balance pitting energy-conserving, short-circuit minimizing, split electron transfer and catalytic speed against damaging oxygen reduction.

Around 1970, semiquinone radical (SQ) was proposed as a key element in mitochondrial respiration with a remarkable role straddling high and low potential redox chains in respiratory Complex III ^1-3. The importance of Complex III in generating reactive oxygen species (ROS) emerged at the same time ⁴. A connection was made between a highly reducing SQ of Complex III and the reduction of molecular oxygen to superoxide radical, the first in a series reactions thought to signal hypoxia and initiate a cascade of events that leads either to cellular protection against low levels of O₂ or cellular destruction in apoptosis ^5-7.

Since then, kinetics and crystal structures have supported the split electron transfer by reduced ubiquinone proposed by Mitchell in 1976 in his Q-cycle mechanism of Complex III ². One electron from hydroquinone at the Qo site passes initially to Rieske FeS cluster, before proceeding down the high potential cofactor chain of heme c₁ and cytochrome c. Essentially simultaneously or, in Mitchell's view after a brief delay in which SQ is an intermediate, the other electron passes first to a low potential heme b_L before going on to a higher potential heme b_H and then on to the Qi site near the other side of the membrane, completing the “Q-cycle” ^{2, 8, 9} (Figure 1a). Extensive studies have made it clear that the Qo site rather than the Qi site dominates ROS generation in hypoxia ⁵.

Figure 1.

a) The redox centres of Complex III are grouped into high and low potential c- and b-chains on opposite sides of the Qo site. When heme b_H is knocked-out a redox quasi-equilibrium is established after a flash. b) Our pH dependence of the redox midpoint potentials in Rb. capsulatus; cyt c₂ is from ³⁰. We estimate the two redox couples of Qo as more oxidizing than cyt c₂, and more reducing than O₂/superoxide ³¹. c) Simulated yield of SQo in heme b_H knock-out mutants following flash activated electron transfer and millisecond redox quasi-equilibrium depends upon pH and the split in the redox couples of the quinone or the commonly used apparent equilibrium stability constant. Using the midpoints of panel b and an optical kinetics determined 1:0.5:1 stoichiometry of RC:cyt c₂:Complex III, a Mathematica calculation of the redistribution of electrons to form the quasi-equilibrium state (in which the redox potential of the low potential chain, including Qo/SQo, and the high potential chain, including SQo/QoH₂, average to the potential of the Q pool) shows that much more SQ is expected after two saturating flashes progressively oxidizes the high potential c-chain (shown here) than after one flash. Similarly, much less SQo is expected if heme b_H is not knocked-out. A yield of 1% at pH 9.0 in the heme b_H knock-out corresponds to an effective K_Stab of ~10⁻¹⁴ to 10⁻¹⁵.

The search for the pivotal SQo began soon after Mitchell's proposal. In 1979 Takamiya failed to detect the SQo signal by EPR in the presence of antimycin ¹⁰ an inhibitor that eliminates the relatively stable semiquinone at Qi. In 1981, de Vries seemed to detect SQo free radical ¹¹, but subsequent work ¹² associated these semiquinones with other respiratory complexes. Recent stopped flow work failed to see any radical at Qo ¹³. SQo was never seen presumably because it is a transient and at most lightly populated intermediate between the first uphill electron transfer to FeS and the second more favourable electron transfer to heme b_L ¹², although others suggested that magnetic coupling between reduced FeS and SQ makes the SQ spin in principle undetectable ¹⁴.

Recent work with cofactor knock-outs trims back the b-chain and provides a unique opportunity to arrest any transient SQ at Qo by hindering escape of the SQ electron. Mutating the heme b_H iron ligating His to Gln in photosynthetic bacteria ¹⁵ creates a Complex III with an intact and fully operational Qo-site ¹⁶ but no heme b_H. A simple flash of light activates the photosynthetic reaction centre (RC) in these membranes, creating in microseconds the two substrates of Complex III: oxidized cytochromes c₂ and hydroquinone. Electron transfer proceeds through Qo along the b- and c-chains until a quasi-equilibrium state is reached, in which the redox potential of the quinone in the Q-pool balances against the average redox potential of the high-potential c-chain and low-potential b-chain ^{2, 16, 17}. The key to promoting semiquinone at Qo is to use the knock-out to create a condition in which the b-chain can only accept one electron in heme bL, but the c-chain can accept two and to use mildly alkaline conditions to lower the redox-midpoint values of the quinone couples 12 to favour oxidation by the high potential c-chain (Figure 1b). Simulations of the electron transfers that lead to the quasi-equilibrium state indicate that SQo should be visible at high pH under a wide range of possible QH2/SQ and SQ/Q redox midpoint values, provided that multiple flashes are delivered to mostly oxidize the c-chain (Figure 1c).

To test this prediction, we redox poise Complex III containing photosynthetic membranes ¹⁶ at the midpoint potential of the Q-pool under anaerobic conditions at pH 9.0, including Complex I and II inhibitors rotenone and carboxin as well as quinol oxidase inhibitor HQNO to minimize possible interfering semiquinone signals. We also include Complex III Qi-site inhibitor antimycin, to eliminate the well-known, relatively stable semiquinone at this site. With or without appropriate Qo-site specific inhibitors stigmatellin or myxothiazol, the material is transferred to an EPR tube ¹⁸, illuminated with from 0 to 8 near saturating flashes of light from a xenon strobe (4 Hz) to initiate electron transfer, followed by rapid freezing in 1 to 2 seconds in a dry ice/acetone bath followed by liquid N₂ storage at least overnight. Figure 2 shows the EPR spectrum of radicals generated by Complex III under these quasi-equilibrium conditions.

Figure 2.

EPR signals at 9.45 GHz, 0.2 mW, 130K of chromatophore membranes exposed to 4 flashes of light (a, c, e) or kept in the dark (b, d, f) for single (a-d) and double (e, f) knock-out mutants with and without Qo-site inhibitors. A stigmatellin sensitive light induced radical is present in heme b_H knockouts at pH 9 (50 mM Tris, 100 mM KCl) (a) but not at pH 6 (50 mM MES) (c) nor in the double b_H/c₁ heme knock-out (e), nor in wild-type with antimycin (not shown). All samples contain 20 μM carboxin and rotenone and 3 fold excess of antimycin and HQNO over approximately 30 μM Complex III as well as 50 μM redox mediators 2,3,5,6-tetramethyl 1,4-phenylenediamine, phenazine methosulfate, phenazine ethosulfate and 2-hydroxynaphthoquinone ¹⁹ to poise at the midpoint potential of the Q pool (-20 mV SHE at pH 9). Stigmatellin or myxothiazol when present is also in three fold excess.

A well-known background radical signal of unknown origin ¹⁹ is present in dark membranes that is unaffected by the presence or absence of a wide range of various Q-site inhibitors (Figure 2b). As predicted, a single flash has essentially no effect on the EPR. However, with 2 flashes the radical signal grows for those samples without Qo-site inhibitors or with the inhibitor myxothiazol, while the stigmatellin samples remain at the background level. By 4 flashes the light-induced radical reaches a maximum and the suppressing effect of stigmatellin is obvious (Figure 2a). As expected for a Qo-site semiquinone, this light-induced signal does not appear at pH 6, where the rise in redox potential of quinone makes oxidation by the flash-oxidized c-chain considerably more unfavourable. Room temperature flash-induced kinetics of EPR concentration chromatophores in thin optical cells confirms that the flash excited RC Q_B-site is not inhibited ²⁰ at the stigmatellin concentrations we use. Optical kinetics also confirm that the flash-induced pseudo-equilibrium dissipates on the ten second time-scale as electrons from the b-chain are reoxidized, perhaps through a slow short-circuit between the b- and c-chains. The light-induced EPR radical signal similarly decays by half on delaying the flash-freeze interval to 5 seconds and decays more or less completely on a 15 second delay.

Further support for the assignment of this light-induced, stigmatellin-sensitive radical to Complex III and not some other adventitiously stigmatellin-sensitive quinone binding complex, we note the stigmatellin-sensitive signal is not seen in antimycin-inhibited wild-type Complex III, where the combined activity of hemes b_L and b_H accommodates two turnovers of the Qo-site. Simulations along the line of Figure 1 predict that under such conditions the available oxidizing power of the high-potential chain is insufficient to accumulate SQ in the Qo-site. Adding a second mutation to the heme b_H knock-out that also effectively knocks-out heme c₁, by replacing its Met ligand with Leu ²¹ and independently slowing electron transfer through the c-chain from milliseconds to seconds, again shows no light-induced radical (Figure 2 e, f). This makes it highly unlikely that the stigmatellin-sensitive signal can be associated with any other quinone reactive protein, such as a quinol oxidase incompletely inhibited by HQNO and the other inhibitors.

The light-induced stigmatellin-sensitive radical signal is distinctly different from other radical signals that can be made in these membranes (Figure 3). The g=2.0040 signal is far from semiquinone Q_A or Q_B signals in native or damaged RC. It is also higher than the 2.0026 signal of the RC bacteriochlorophyll anion (P⁺), which can be generated in these samples by illuminating them at ~130K in the EPR cavity. Even though this signal shares a similar ~0.7 mW power saturation as SQi, it has a conspicuously lower g-value. The signal is also noticeably broader, with a 11.7 gauss splitting between the positive and negative peaks, compared to 8.5 gauss for SQi. We conclude that the most likely explanation of our observations is that at pH 9.0, a light-induced quasi-equilibrium lasting less than 15 seconds in the heme b_H knockout generates a SQ at Qo which can be prevented by addition of the Qo-site inhibitor stigmatellin. Compared to SQi, the g-value of the stigmatellin-sensitive radical suggests a more polar and/or more strongly hydrogen bonding environment while its greater width may indicate some spin interaction with the nearby Rieske FeS centre.

Figure 3.

A comparison of light-induced radical EPR signals in photosynthetic membranes: a) the g-value associated with the noticeably wide, light-induced-stigmatellin sensitive signal (red) is in between that of the light-induced antimyc in-sensitive SQi anion (blue) and the low temperature light-induced bacteriochlorophyll-dimer radical cation (green, shown at 1/50 amplitude, and generated with continuous illumination from a 150 W halogen lamp through a fiber optic projected into the EPR cavity); however, the apparent SQo signal cannot be approximated by any sum of SQi and P⁺. b) Power saturation of the stigmatellin-sensitive light-induced radical is comparable to that of SQi but at significantly lower power than P⁺.

Spin quantitation using the in situ generated P⁺ signal together with the multi-flash optical kinetics generated 1:1 stoichiometry of Complex III to RC in the heme b_H knock-out, indicates that we can generate SQ in about 1% of the Complex III, assuming there is no spin-spin interactions. This yield corresponds to an ~880 mV split in the redox midpoints of the two Qo semiquinone couples (Figure 1c), placing the Qo-site semiquinone/hydroquinone couple between the cyt c₂ and P⁺ couples. This is on the low end of the effective stability constant range that will still support sub-millisecond electron tunnelling and catalysis given the likely electron transfer distances in the Qo-site ²². It appears that the SQo redox properties may have been naturally selected to be to keep concentrations as low as possible to suppress superoxide generation from O₂ but still high enough to enable rapid, productive electron transfer. SQo apparently becomes conspicuous when the high-potential chain is highly oxidized and the low-potential chain highly reduced, i.e. conditions in which ROS was originally observed ⁴.

One unexpected observation is that the Qo-site inhibitor myxothiazol does not impede the creation of the light-induced radical as effectively as stigmatellin (Figure 2a). Myxothiazol and other methoxyacrylate inhibitors such as MOA-stilbene bind to a distinctly different domain than stigmatellin, on the opposite, b heme side of the Qo-site ^{23, 24}. Stigmatellin displaces quinone and removes the normal sensitivity of the Rieske EPR to Q-pool size. However, with MOA-stilbene, the Rieske EPR signal remains sensitive to Q-pool size, suggesting that some quinone may still approach the Qo-site, even though normal catalytic turnover is inhibited ^{23, 24}. Several methoxyacrylate inhibitors bind non-competitively with quinone, also suggesting that the inhibitor and quinone can bind simultaneously ²⁵. If myxothiazol does allow quinone binding and some SQo formation as our results suggest, this would rationalize the observation that myxothiazol permits considerably more superoxide production by Complex III than stigmatellin ^26-29.

Although we have good evidence that we can generate semiquinone at Qo, it is not yet proven that this semiquinone is the reductant for molecular oxygen, despite as yet circumstantial evidence that rapid addition of oxygen to our samples during illumination before rapid freezing decreases the light-induced radical signal. Nor is it yet proven that SQ is a genuine intermediate in normal Qo site action, as expected from a Mitchell-like sequential electron transfer mechanism as opposed to a genuinely concerted electron transfer mechanism ⁹.

Nevertheless, detection of this species finally makes direct tests of these thirty year old models feasible and illuminates the essential natural engineering balance that may drive the selection of the Qo site quinone thermodynamic and electron tunnelling properties. In a sequential electron transfer, a small split between the Qo site quinone couples should lead to relatively slow endergonic heme b_L electron transfer, while increasing the split, dropping the reducing quinone couple to or below the similar heme b_L and superoxide redox potentials, speeds both these electron transfers. Corresponding electron transfer to FeS slows, which helps minimize intermediate SQ availability for electron transfer with O₂, but tends to impede bio-energetically essential catalytic turnover. This cannot be overcome simply by securing the FeS and heme b redox centres closer to the quinone and each other, as this would speed the energy wasting short-circuit electron transfer between heme b_L and FeS. The natural engineering design compromise of Qo appears to be: 1) place the heme b_L and FeS redox partners of Qo more than 20 Å apart to slow direct short-circuits to seconds; 2) place the quinone nearly in line between the two centres, to speed productive electron tunnelling times; 3) split the Qo redox couples so that electron transfer to FeS is significantly uphill, so that SQ is only briefly available to produce superoxide, but split no further than the ultimate source of oxidants--photosynthetic reaction centres or respiratory oxygen reducing centres; 4) avoid excessive slowing of uphill FeS electron transfer and catalytic turnover by keeping Qo closer to FeS than heme b_L and provide proton transfer partners so that uphill electron transfer proceeds at nearly the electron tunnelling rate.