Hydrogen bonding rearrangement by a mitochondrial disease mutation in cytochrome bc 1 perturbs heme b H redox potential and spin state.

Hemes are common elements of biological redox cofactor chains involved in rapid electron transfer. While the redox properties of hemes and the stability of the spin state are recognized as key determinants of their function, understanding the molecular basis of control of these properties is challenging. Here, benefiting from the effects of one mitochondrial disease-related point mutation in cytochrome b, we identify a dual role of hydrogen bonding (H-bond) to the propionate group of heme b H of cytochrome bc 1, a common component of energy-conserving systems. We found that replacing conserved glycine with serine in the vicinity of heme b H caused stabilization of this bond, which not only increased the redox potential of the heme but also induced structural and energetic changes in interactions between Fe ion and axial histidine ligands. The latter led to a reversible spin conversion of the oxidized Fe from 1/2 to 5/2, an effect that potentially reduces the electron transfer rate between the heme and its redox partners. We thus propose that H-bond to the propionate group and heme-protein packing contribute to the fine-tuning of the redox potential of heme and maintaining its proper spin state. A subtle balance is needed between these two contributions: While increasing the H-bond stability raises the heme potential, the extent of increase must be limited to maintain the low spin and diamagnetic form of heme. This principle might apply to other native heme proteins and can be exploited in engineering of artificial heme-containing protein maquettes.

Hemes are common redox-active cofactors in biological electron transfer systems. Their major function is to transfer electrons within the cofactor chains or as part of the catalytic sites. The direction and rate of electron transfer are secured by the specific properties of hemes, among which the redox midpoint potential and the spin state are considered to be of crucial importance. Several studies have been carried out to understand how specific molecular elements contribute in adjusting the redox potential values to the levels required for efficient electron transfer rate (1–5). These studies recognized the importance of the heme-iron ligands, the degree of exposure of the heme to the solvent, and specific interactions, including hydrogen bonding, within the amino acids of heme-binding pocket. However, the overall structural complexity of proteins often makes an experimental extraction of individual elements of control extremely challenging, in particular those associated within the hydrogen bonding networks. It is indeed remarkable that the fluctuation of the hydrogen bond network was identified as the factor modulating the efficiency of long-range biological electron transfer (6, 7).

This work focused on exploring the hydrogen bond-related elements of control of the redox properties of one of the hemes of cytochrome b subunit of cytochrome bc1 (mitochondrial complex III). Cytochrome b is the only subunit of this complex encoded by mitochondrial DNA, thus is a subject of high susceptibility for spontaneous mutations in comparison to other subunits (encoded by nuclear DNA). Such mutations often lead to systemic disorders, mitochondrial diseases, manifesting in symptoms such as exercise intolerance, myopathies, or neuropathies. On the other hand, their mimic in the bacterial or yeast model systems not only provides insights into the possible molecular basis of the disease but also reveal molecular aspects of protein design (8, 9). With this in mind, we targeted a mutation G34S (human cytochrome b numbering), found in one 52-y-old female patient suffering from exercise intolerance (10). The mutation was present in mitochondria from muscle tissue and caused mild myopathy, lactic acidosis, and defect in mitochondrial complex III activity. The mutated glycine 34 is located on transmembrane helix A of cytochrome b subunit in close distance (∼8 Å) to heme bH (). This location suggests the possible effect of mutation on the properties of heme cofactor, in particular if one considers that a change of small and nonpolar glycine to a larger and polar serine may affect the local hydrogen bonding environment.

Indeed, the presence of glycine at position 34 was previously noticed as important for heme packing in the cytochrome b subunit (11). Interestingly, the importance of G34 was emphasized by the remarkable evolutionary conservation in the equivalents of cytochrome bc1 complex in distant organisms, such as bacteria, insects, fish, mammals and even plants (12). In a bacterial model (Rhodobacter sphaeroides cytochrome bc1), the equivalent G48 residue was mutated to valine and aspartic acid, both rendering the bacteria photosynthetically incompetent (13). The equivalent of this mutation was also studied in Saccharomyces cerevisiae (8). Yeast with G33S (S. cerevisiae numbering) mutation were not able to grow aerobically, whereas the isolated bc1 complex had lower enzymatic activity and disrupted subunit composition (lower level of the iron-sulfur protein assessed by Western blotting and lower cytochrome b content, measured optically to 55% of wild type [WT]). The same G33 position was found to be mutated to aspartic acid in a respiratory-deficient yeast (14). Among the 85 tested G33D revertants, 82 were D33G and 3 were D33A, suggesting that only a small amino acid without a charge can be tolerated at this position.

Given the implicated importance of glycine at 34 position of mitochondrial cytochrome b, particularly in the context of hydrogen bonding network within the heme-binding pocket, we combined the experimental and computational methods to investigate the effects of introducing serine to the homologous position in purple photosynthetic bacterium Rhodobacter capsulatus (mutant G48S). Notably, the bacterial cytochrome b shares reasonable sequence similarity with human and yeast mitochondrial cytochrome b, close to 57%. We found that G48S perturbs the redox properties of heme bH and causes structural and hydration changes in the vicinity of heme.

The mutation also severely affects the spectral properties of heme bH, which are best explained by a model assuming a reversible change from low to high spin state when the heme is oxidized. The latter comes as a rather unusual and unexpected molecular effect, considering all so-far reported effects of disease-related mitochondrial mutations (8, 9, 15). At the same time, it offers interesting insights into the role of hydrogen bonding and protein packing in maintaining the low spin state of the oxidized heme fostering electron-transfer relay function.

Results

The Spectroscopic Properties of G48S Mutant.

The unusual electron paramagnetic resonance spectra of hemes b in G48S mutant.

As a first step of investigation of the molecular effects of mitochondrial mutation equivalent in R. capsulatus, the basic characterization of the mutant was performed. Bacteria with G48S mutation on cytochrome b in comparison to WT exhibited only slightly slower growth under cytochrome bc1-dependent photosynthetic conditions, suggesting that mutated complex is functional (). However, several attempts of G48S purification resulted in varying stoichiometry of the complex subunits indicating its low stability in solution. Therefore, all of the further analyses were carried out using membrane chromatophores. The optical absorption spectra of hemes and electron paramagnetic resonance (EPR) spectra of [2Fe-2S] cluster did not reveal significant changes in G48S chromatophores in comparison to WT (). Also, the light-activated kinetics (measured under conditions with quinone [Q] pool half-reduced) did not reveal any major difference in rates of heme bH reduction from quinol oxidation at Qo site, with only heme bH reoxidaton impaired in G48S in comparison to WT (Table 1).

Table 1.

Selected properties of G48S mutant in comparison to WT

	Phenotype*	Light-induced heme b†		Qi reverse	Em7 of hemes			Em7 of hemes
		Reduction, s−1	Reoxidation, s−1	Reaction, s−1‡	Spectrophotometry			EPR
					bHH, mV	bHL, mV	bL, mV	bHL, mV	bL, mV
WT	+++		180	400	128	60	−124	nd§	nd
+ ant		1,050			—	60	−120	51	−102
G48S	++		50	35	148	21	−136	—	—
+ ant		980			—	73	−112	62	−105

++ indicates photoheterotrophic growth slower than WT.

Measured at pH 7.0 and Eh = 100 mV (Q pool half-reduced).

Measured at pH 9.0 and Eh = 250 mV (Q pool oxidized).

nd, not determined.

However, the unusual behavior of hemes in G48S was revealed by continuous wave (CW) EPR spectroscopy at low temperature. Under typical conditions, the WT sample fully oxidized with ferricyanide exhibits the g = 3.41 transition, corresponding to oxidized low spin heme bH (16). This signal is also present in the ascorbate-reduced sample (Fig. 1, black), indicating that the heme remains oxidized, which is consistent with the well-described low redox midpoint potential (Em7) of this heme (∼50 mV) (17). In G48S mutant, this EPR transition was not present neither in a sample fully oxidized (see top black spectrum in Fig. 2, Right) nor reduced with ascorbate (Fig. 1, black). At the same time, the transition corresponding to oxidized heme bL was observed in G48S with the same value g = 3.78 as for WT sample (Fig. 1). In previous studies, the lack of g = 3.41 was observed when the bis-His ligation pattern of heme bH was changed by mutation (18). Thus, the lack of g = 3.41 in G48S immediately suggests that this mutation also might introduce structural changes affecting the heme-ligand interactions.

Fig. 1.

Comparison of selected spectroscopic properties of WT and G48S in chromatophores. CW EPR spectra of hemes: (A) WT sample reduced with ascorbate (black) and with subsequent inhibition by antimycin (red); (B) G48S sample reduced with ascorbate (black) and subsequent inhibition by antimycin (red); and (C) G48S oxidized sample inhibited with antimycin (red) and with subsequent reduction by ascorbate (green). Optical spectroscopy (D and E): light-induced redox kinetics of hemes bH for WT and G48S, respectively. Reduction of heme bH through reverse electron transfer at the Qi site followed at 560 to 570 nm. Traces were recorded without inhibitors (black), with myxothiazol (blue), and with myxothiazol and antimycin (red) at pH 9 and ambient potential of 250 mV; (F and G) optical redox titration of b-type hemes measured for WT and G48S chromatophores respectively at pH 7.0 for samples without inhibitors (black) and with antimycin (red).

Fig. 2.

CW EPR spectral changes in G48S chromatophores upon reductive and oxidative redox titrations. Reaction sequence: 1) reductive titration from fully oxidized sample to value below the Em7 of WT heme bH; 2) addition of antimycin (ANT); and 3) oxidative titration rising the potential to the initial value. (A) Example of CW EPR spectrum of G48S with marked regions of high and low spin resonance transition of hemes. (B) Parts of CW EPR spectra showing transitions g = 5.99 (Left) and g = 3.40 (Right) recorded at various redox potentials. (C) The dependence of CW EPR signal amplitude of high spin (Left) and low spin (Right) on the ambient redox potential (Eh) obtained from the spectra shown in B; the color of dots indicates reductive and oxidative potentiometric titration (black and red, respectively).

Interestingly, addition of antimycin (the Qi site-specific inhibitor) to the ascorbate-reduced sample resulted in the appearance of resonance transition g = 3.40 reflecting the oxidized heme bH (Fig. 1, red). This phenomenon was not observed for the fully oxidized sample inhibited with antimycin (Fig. 1, red); to observe the g = 3.40 transition, the G48S sample must have been reduced with ascorbate prior to the addition of antimycin (Fig. 1, green). It is of note that the g value (g = 3.40) of G48S corresponded to the g value of heme bH in WT in the absence of antimycin (in WT, addition of antimycin results in the shift of its EPR g transition from 3.41 to 3.45).

These results indicate that in G48S mutant, the redox properties of heme bH are significantly changed in comparison to WT. The most straightforward explanation for the lack of g = 3.41 EPR transition in the G48S sample is an increase in the Em of heme bH, which would have to be large enough to render this heme fully reduced in the presence of ascorbate (i.e., the Em of the heme bH would have to be in the range of 200 mV or more). This prompted us to look more closely into the redox properties of this heme.

Influence of G48S mutation on the redox properties of heme bH observed by optical spectroscopy.

The redox equilibrium between hemes b and the Q pool was monitored by light-induced electron transfer under the conditions favoring electron transfer from quinol occupying Qi site to oxidized heme bH. This reaction can be observed with the Qo site-specific inhibitor (in this work, myxothiazol) when the reducing power of the Q pool is increased (by increasing pH) and the Q pool is oxidized before the light activation (19). The rate of heme bH reduction under these conditions for G48S mutant is approximately 10-fold lower than WT (Table 1), whereas the amplitude is two times higher (Fig. 1 , blue) approaching the level of heme bH reduction in the absence of inhibitors (Fig. 1 , black). The observed increase in the amplitude of heme bH reduction in G48S could be a result of either an increase in the Em of the heme or a modified interaction of the heme with the redox-active occupant of the Qi site, which would lead to an “apparent” increase of Em of the heme.

To distinguish between these two possibilities, the optical redox potentiometric titrations of the G48S mutant in the absence and presence of antimycin were analyzed in reference to WT. At pH 7, hemes b in WT without inhibitors titrate with Em values of +128 mV, +60 mV, and −124 mV, corresponding to heme bHH (a fraction of heme bH with higher Em), heme bHL (a fraction of heme bH with lower Em), and heme bL, respectively (Fig. 1, black). The presence of two forms of heme bH (bHH and bHL) is well established and interpreted as an indication of interaction of the heme with the redox-active occupant of the Qi site (20, 21). When this interaction is prevented in the presence of antimycin, heme bH titrates as one component with Em7 of +60 mV (Fig. 1, red).

G48S without inhibitors titrates with the three components having Em values of +148 mV, +21 mV, and −136 mV (Fig. 1, black), which correspond to hemes bHH, bHL, and bL of WT, respectively. The main difference is a considerably larger fraction of the high-potential heme bHH (Em +148 mV), which in G48S accounts for about 50% of total b-type hemes, whereas it only accounts for 30% in WT. In the presence of antimycin, heme bH in G48S titrates as one component with Em of +73 mV (Fig. 1, red), which closely resemble titrations of this heme in WT. The larger amplitude of heme bHH in G48S in the absence of inhibitors and the lack of significant changes in Em of this heme in the presence of antimycin indicate that the apparent increase in Em of heme bH in G48S is in part related to modified interaction of this heme with surroundings and also in part with the redox-active occupant of the Qi site.

Intriguingly, the magnitude of this increase is not large enough to explain the lack of g = 3.40 transition in the ascorbate-reduced sample in the absence of antimycin (Fig. 1, black). To solve this discrepancy, we performed EPR redox titrations of hemes b in G48S and WT. These titrations monitor disappearance/appearance of oxidized low spin hemes and thus complement the optical titrations, which monitor disappearance/appearance of the reduced hemes. We thus expected that combining both types of titrations would help us understand the properties of heme bH in G48S.

Indication of heme bH spin state conversion by G48S mutation from EPR measurements.

The EPR redox titrations were carried out in a wide range of magnetic induction to observe changes of all possible resonance transitions upon changing the redox state of the sample. The reductive titration of fully oxidized sample did not reveal any signal of low spin paramagnetic centers (highly axial low spin [HALS] type) that could be assigned as heme bH (Fig. 2, step 1). Moreover, addition of antimycin at ambient redox potential (Eh) ∼40 mV (which is below the Em value of heme bH in WT) did not change the EPR spectra of hemes (Fig. 2, step 2). However, the subsequent oxidative titration with ferricyanide (i.e., after addition of antimycin) resulted in the appearance of resonance transition g = 3.40, the amplitude of which increased with the increase of Eh reaching a maximum at around 200 mV (Fig. 2, step 3). The Em of this transition was calculated to be +62 mV at pH 7.0 (Fig. 2, Right), which is remarkably similar to the Em of heme bH described for WT (17). Given this similarity and the position of g, this transition was assigned as the low spin heme bH (22, 23).

Interestingly, along with the changes seen in g = 3.40 transition, the unusual and absent in WT changes of resonance transition g = 5.99 were also observed. During the reductive titration, the amplitude of this transition decreased significantly (Fig. 2, green dashed line) and did not rise to the initial value during the oxidative titration (Fig. 2, Left). Considering that g = 5.99 transition is characteristic for high spin oxidized hemes (23, 24), it suggests that in contrast to WT in mutant, there is a fraction of high spin hemes. During the reductive titration of G48S, the low spin heme bH signal is not observed but appears during the oxidative titration phase. It is thus clear that the g = 3.40 transition appears at the expense of g = 5.99. To explain all these observations, we conclude that the oxidized heme bH in the G48S is high spin (no g = 3.40 transition) and upon reduction undergoes conversion to low spin (no g = 3.40 transition but optical maximum at 560 nm).

The interactions of G48S at atomic level.

All experimental results indicated some prominent changes of spectroscopic and redox properties caused by G48S mutation. However, their molecular bases were difficult to explain. Thus, we engaged computational methods to study the possible structural changes and interactions of mutated residue with local protein environment, especially with heme bH. To investigate the effects of mutation at an atomic level, classical atomistic molecular dynamics (MD) simulations were performed on models built from crystal structures of R. capsulatus and S. cerevisiae cytochrome bc1 (Protein Data Bank [PDB] IDs 1ZRT and 3CX5, respectively) (25, 26). Additionally, in order to shed light on the redox behavior of heme bH, density functional theory (DFT) calculations on WT and mutant cluster models were performed.

Structural rearrangement and altered hydrogen bonding caused by G48S.

The results of MD simulations reveal a clear local difference between the WT and G48S mutant. In bacterial WT complex simulations, the backbone of Gly48 (-NH) makes an H-bond with the propionate group of heme bH with 28% ± 4% occupancy of all simulated trajectories (Fig. 3). However, in mutant, structural rearrangement takes place, causing H-bond formation between heme bH propionate and Ser48 sidechain (-OH) instead of the protein backbone. This relatively stable H-bond occurs rapidly (<1 ns) with an occupancy of 53% ± 6% over all G48S simulation trajectories (Fig. 3). The same scenario is observed in MD simulations of mitochondrial cytochrome bc1, where even more drastic difference occurs in H-bond occupancies of WT and mutant. While the serine sidechain-based H-bond to propionate is highly stable for 79% ± 1% of total simulation time, the glycine backbone forms a weaker H-bond with ca. 13% ± 2% occupancy (). Interestingly, presence of a bulkier serine instead of glycine also causes structural perturbation, which is observed as slight bending of the transmembrane A helix (Fig. 3). In addition, two different conformational arrangements are observed in simulations of mutant bacterial cytochrome bc1. In one case, Ser48 resides next to the His111 ligand of heme bH (“near” arrangement, ca. 6 Å distance between Ser Cα and Fe atoms), whereas in another case it shifts further away (“far” arrangement, ∼8.5 Å distance between Ser Cα and Fe atoms), with the H-bond to propionate still retained (Fig. 3). In contrast, the more tightly packed “near” arrangement is found to be highly stable in WT Gly case (Fig. 3). This tight packing partly restricts the conformational freedom of the His111 ligand (). We also find that due to the larger structural variability of polar side chain of serine as in “far” arrangement, more space is created, which also results in a higher level of hydration near the heme propionate group in mutant in comparison to WT enzyme (Fig. 3 ). Clearly, mutation of Gly to Ser changes the dynamics and polarity of the region, which may explain the spectroscopic differences observed between WT and mutant.

Fig. 3.

Snapshots from MD simulation of bacterial cytochrome bc1 showing H-bonding between propionate and Gly backbone in WT (A) and Ser sidechain in mutant (B). The hydrogen bond occupancy is labeled. (C and D) The structural rearrangement of transmembrane helix A in WT (blue) and G48S (red). The two conformations of Gly and Ser (“near” and “far”) are shown as a histogram (normalized such that area under the curve is equal to one) of distance between backbone Cα of Gly/Ser and Fe of heme bH. (E and F) Water occupancy near the propionate region in WT and G48S mutant, respectively. Water occupancy is shown as dotted surface (20% occupancy, data from all simulation trajectories). (G) Simulation snapshot showing H-bond between propionate and Gly backbone in WT with antimycin (Ant) bound the at Qi site. (H) Simulation snapshot with H-bonding between G48S sidechain and antimycin is shown, together with the H-bond occupancies.

The results from DFT calculations on bacterial model systems show that when Ser sidechain is H-bonding to the propionate group, as opposed to Gly backbone, it enhances the electron affinity of heme bH by ∼0.13 eV. This effect is observed in all three different DFT functionals used (0.11 ± 0.05 eV in B3LYP, 0.15 ± 0.06 eV in TPSS, and 0.14 ± 0.07 in TPSSh). The increase in electron affinity is also seen based on DFT calculations performed on high-resolution structure of mitochondrial bc1 complex, albeit the effect is relatively milder (30 meV to 40 meV across the three density functionals utilized). Overall, the data clearly indicate that even the subtle change in hydrogen bonding environment affects the heme bH redox properties.

The influence of antimycin on H-bonding arrangement in G48S mutant.

In order to study the effects of antimycin on H-bond dynamics, we also performed MD simulations on an antimycin-bound structure of bacterial complex. Due to relatively weaker stability of antimycin in our MD simulations, we restrained it to its crystallographic position by harmonic restraints (see Table 2). We found that Ser still makes H-bond to propionate (∼20% ± 5% of total simulation time). However, stable H-bonds to the carbonyl group of antimycin are observed (∼30% ± 9% occupancy) as long as it binds to the Qi site in a stable crystallographic arrangement (Fig. 3). Interestingly, the H-bond to antimycin from Ser occurs on an exclusive basis such that H-bond to propionate ceases to exist. In WT case, a weak H-bond to propionate was formed for ∼15% ± 5% of the simulation time; however, the Gly backbone was unable to form the hydrogen bond with antimycin. Overall, the antimycin-bound WT state resembles the H-bonding situation in the quinone-bound WT state, but mutant explores an alternative H-bonding scenario with antimycin. In addition, the “near/far” dynamics of antimycin-bound cytochrome bc1 are found to be similar to quinone-bound simulations (both WT and mutant; Fig. 3).

Table 2.

Details of the MD simulation setups (B1-B4 bacterial bc1 complex and M1-M2 mitochondrial bc1 complex)

Setups	WT/mutant	Qo/Qi occupation	Replicas × time (ns)
B1	WT G48	Quinone/quinone	3 × 400 ns
B2	Mutant G48S	Quinone/quinone	3 × 400 ns
B3	WT G48	Quinone/antimycin*	3 × 300 ns
B4	Mutant G48S	Quinone/antimycin*	3 × 300 ns
M1	WT G48	Quinone/quinone	3 × 300 ns
M2	Mutant G48S	Quinone/quinone	3 × 300 ns

Antimycin was restrained using harmonic spring to keep the head group in place, using a force constant of 1,000 kJ ⋅ mol−1 ⋅ nm−2.

Discussion

Spectroscopic Studies and MD Simulations Implicate That H-Bonding to Heme Propionate Affects the Redox Potential of Heme bH.

Data from our classical MD simulations reveal distinct differences between WT and mutant cases and allow us to provide a molecular interpretation to experimental observations. We suggest that the high potential heme bHH signal is in part caused by the H-bond to propionate from Gly or Ser. Indeed, our DFT calculations support that a Ser based H-bond may enhance electron affinity of heme bH. The higher proportion of heme bHH in the mutant seen in experiments is in excellent agreement with the higher occupancy of the Ser-propionate H-bond compared to Gly. The heme bHL can therefore be assigned to the state where there is no hydrogen bond between propionate and residue in concern. MD simulations also suggest that a tightly bound antimycin can steal the hydrogen bond with Ser to the propionate. This also agrees with the spectroscopic data, which saw a loss of heme bHH in the mutant with antimycin present. Overall, our atomistic simulations and DFT calculations provide a molecular picture on the origin of heme bH signals observed spectroscopically and discussed recently in ref. 21.

Model of Spin Conversion Observed for G48S Mutant.

On the basis of experimental data and supporting results from MD simulations, the following molecular model of redox and magnetic states of heme bH in G48S can be proposed (Fig. 4). As this model considers changes in the spin state of the heme, we first describe a well-established case of the WT heme bH (16). The fully oxidized heme bH has low spin (S = 1/2) and therefore can be observed in CW EPR with characteristic transition g = 3.41. Optical spectrum of oxidized heme in “α” region is broad with no absorption maximum at 560 nm. Upon reduction, the spin state of heme bH changes to S = 0, which renders this heme invisible by EPR spectroscopy. The reduction is associated with appearance of absorbance maximum at 560 nm in optical spectroscopy. Moreover, addition of Qi site inhibitor, antimycin, at any redox state does not affect the spin state of heme bH (Fig. 4).

Fig. 4.

Model of the spin state conversion of heme bH in G48S. (A) In WT, the oxidized heme bH is low spin S = 1/2 (characteristic transition gx = 3.41 in CW EPR spectroscopy; no maximum absorption at 560 nm in optical spectroscopy). Upon reduction, the heme bH becomes diamagnetic and invisible in CW EPR (spin state S = 0) but can be observed on optical spectrum with maximum absorbance at 560 nm. Addition of antimycin does not affect the heme bH spin state. (B) In G48S, the oxidized heme bH is high spin S = 5/2 (no EPR signal at g = 3.40 and no maximum at 560 nm in optical spectroscopy). The addition of antimycin to G48S in the oxidized state does not change the spin state of heme or the appearance of EPR and optical signal. However, the prereduction of G48S sample with sodium ascorbate leads to spin conversion of heme bH (from S = 5/2 to S = 0). Then, the reduced heme bH is observed only in optical spectroscopy (no EPR signal). The addition of antimycin leads to lowering of heme bH midpoint potential and in consequence oxidizing the sample prereduced with ascorbate (Fig. 1), which then could be observed in EPR as the appearance of g = 3.40 transition (Fig. 2).

In G48S, in contrast to WT, the heme bH spin state depends on its redox state. The fully oxidized heme bH has high spin (S = 5/2), which results in the lack of the characteristic signal g = 3.40 in CW EPR. This high spin heme has a transition at g = 5.99, which in membrane samples overlaps with other signals but can be distinguished upon careful analysis of EPR redox titrations (Fig. 2). The addition of antimycin to oxidized heme bH in G48S does not change its spin state, and consequently, the g = 3.40 transition is absent (Fig. 1, red). The reduction of this heme leads to the spin conversion from S = 5/2 to S = 0. Such form has a clear maximum in optical spectrum (560 nm) and still no g = 3.40 transition. Only after subsequent oxidation does the g = 3.40 transition appear in the EPR spectrum of this mutant, meaning that this heme achieves low spin state (S = 1/2).

Simulation analysis shows that in WT enzyme “near” arrangement is predominantly stabilized (Fig. 3), which partly restricts the conformational flexibility of histidine ligand (). This is most likely due to the smaller volume of Gly residue that allows tighter packing in the vicinity of heme and its ligand. A glycine-induced cramped region would prevent flexibility of histidine ligand and thus help in stabilizing low spin state of heme. In contrast, population of “far” arrangement is much more enhanced in G48S mutant caused by the sidechain of serine. Such a change provides additional freedom to histidine ligand (His-111) of heme bH to relax to higher spin states. Furthermore, due to the space created upon replacing Gly with Ser, hydration is enhanced in the vicinity of the heme propionate, which may also increase redox potential of the heme (see also ref. 27).

With this model, we can explain the ensemble of optical and EPR redox titrations (Figs. 1 and 2) as well as the spectral changes in G48S relevant to the sequence of events described in Fig. 1 . For the case shown in Fig. 1, the addition of antimycin to ascorbate-reduced sample lowers the heme bH midpoint potential, resulting in its oxidation. This makes the heme bH detectable in EPR as g = 3.40 transition (red spectrum), which could not be observed before the spin conversion (black spectrum). The case shown in Fig. 1 can be explained with two possible scenarios. In the first one, antimycin does not bind to oxidized high spin heme bH until this heme is reduced by ascorbate. After reduction, the heme undergoes conversion to low spin and then antimycin binds, lowering its Em, which results in oxidation of this heme. In the second scenario, antimycin binds to oxidized high spin heme bH and lowers its Em. Addition of ascorbate induces spin conversion of this heme but does not reduce it. Because both of these scenarios would result in the appearance of EPR g = 3.40 transition, typical of oxidized low spin heme (as seen by comparison of the red vs. green spectrum in Fig. 1), it cannot be decided whether antimycin binds to the oxidized high spin or to reduced low spin heme bH.

The proposed model provides a detailed understanding of the molecular effects of exercise intolerance-related homologous Gly to Ser mutation in cytochrome b of human mitochondrial complex III. Given the light-induced electron transfer competence of G48S substitution (Table 1), one would expect that the overall electron flux is not largely altered in the mutant cells. Nevertheless, the heme b reoxidation and the Qi site reverse reaction are slowed by 4- and 10-fold, respectively, in the mutant in comparison to the WT. The spin conversion in G48S might be at the root of these perturbed kinetic effects. MD simulations indicate that H-bonding patterns of heme bH propionate in bacterial and mitochondrial Gly to Ser mutants are similar. Given the overall similarities, also at the sequence level, it follows that heme bH in mitochondrial mutant might also undergo spin conversion, which, in analogy to the bacterial system, might introduce limitations to the rate of electron transfer. One may expect such limitations to have a larger impact on the cells especially under stress or the conditions of increased energetic demands (high proton motive force, for instance). Another possible molecular effect that contributes to the exercise intolerance is the decreased stability of the mutated cytochrome bc1 complex implicated from this study and also from the studies with yeast cytochrome bc1 (8). We note that overall effects of this Gly to Ser mutation are expected to be subtle, as reported in other cases (see example in ref. 28); otherwise, the mutation would have been lethal to the cells.

The Dual Effect of the Strength of H-Bond to Heme Propionate.

Considering mutational studies on other redox-active enzymes, it is reasonable to expect that the alteration of H-bonds to heme propionates modulates the redox potential of the heme (see examples in refs. 29–31). In view of our data, the H-bond to heme propionate from the amino acid at position 48 appears to have a dual effect. On one hand, change of H-bond donor from backbone of Gly to sidechain of Ser leads to its stabilization and an increase in the redox potential of the heme. On the other hand, it appears to perturb interactions between Fe ion and axial histidine ligands, which leads to a reversible spin conversion of the oxidized Fe ion from S = 1/2 to S = 5/2. However, the latter effect should be regarded as undesired for hemes that are part of electron transfer chains. The spin state of those hemes is typically S = 0 and S = 1/2 for reduced and oxidized state, respectively, and spin conversion at a given redox state is avoided. This is because such a conversion can potentially reduce the electron transfer rate between any given heme and its redox partners, as implicated from kinetic impediments between heme bH and the occupant of the Qi site in G48S. In view of this dual effect, it can be suggested that the strength of the H-bond in the native protein was optimized to secure both the high potential and the low spin and diamagnetic form of heme in the oxidized and reduced state, respectively. We anticipate that similar design principles might apply for other native heme proteins and, possibly, be exploited when designing artificial heme protein maquettes of specific redox properties (32, 33).

Materials and Methods

Genetic Modifications.

R. capsulatus cells containing glycine to serine substitution at position 48 in cytochrome b subunit were obtained using a genetic system described in ref. 34. G48S mutation was introduced in the petB gene coding for cytochrome b gene using QuikChange site-directed mutagenesis system (Stratagene) and the following PCR primers (the mutated nucleotides are marked in bold in the underlined triplet): G48S_FWD, 5′–GAA​CTG​GTG​GTG​GAT​CTG​GGATC​GTC​CTT​GCC​TTC​ACG–3′, and G48S_REV, 5′–GTG​AAG​GCA​AGG​ACG​ATCCCA​GAT​CCA​CCA​CCA​GTT​C–3′. As a template DNA, the pPET1 plasmid containing WT petABC operon was used. The BstXI-AsuII fragment of the operon containing the desired mutation, and no other mutations, was inserted into pMTS1 vector and introduced into MT-RBC1 R. capsulatus strain using triparental crossing (34). The presence of introduced mutation was confirmed by sequence analysis of petB gene on plasmid isolated from mutated R. capsulatus strain.

Cell Growth and Preparation of Chromatophores.

R. capsulatus bacteria were grown under semiaerobic conditions on mineral-peptone-yeast extract (MPYE) medium. The tests of bc1-dependent growth under photosynthetic conditions were performed on MPYE-agar plates containing 100 μL bacteria diluted from overnight cultures to OD600 = 0.1, incubated in anaerobic jars (GasPak System, BD) at 30 °C under continuous light. Chromatophore membranes were prepared from 10 L liquid cultures with a French pressure cell at 16,000 psi followed by separation on ultracentrifuge according to the procedure described previously (35). After homogenization chromatophores were suspended in 50 mM 3-(N-Morpholino)propanesulfonic acid (MOPS) buffer (pH 7.0, 100 mM KCl, 1 mM EDTA) and stored at 4 °C.

Light-Activated Redox Reactions of Hemes.

Double-wavelength time-resolved spectrophotometer (36) was used to measure the kinetics of electron transfer through the heme cofactors of cytochrome bc1. Measurements were performed according to refs. 37 and 38 using chromatophores suspended in MOPS buffer poised at redox potential 100 mV or in 50 mM Tris buffer (pH 9.0, 100 mM KCl, 1 mM EDTA) poised at 250 mV. Sodium dithionite or ferricyanide were used to adjust the ambient redox potential (Eh) of chromatophore suspensions. The samples were sustained under anaerobic conditions in the presence of 3.5 μM valinomycin and the following redox mediators: 7 μM 2,3,5,6-tetramethyl-1,4-phenylenediamine, 1 μM phenazine methosulfate, 1 μM phenazine ethosulfate, 5.5 μM 1,2-naphthoquinone, and 5.5 μM 2-hydroxy-1,4-naphthoquinone. Transient kinetics of hemes b were monitored at 560 to 570 nm after activation by single saturating light pulse (∼10 μs). The rates of flash-induced electron transfer reactions were determined from single exponential function fitted to kinetic traces of heme bH reduction (in the presence of antimycin), bH reoxidation (without inhibitors), and heme bH reduction from reverse reaction (in the presence of myxothiazol). Inhibitors were used at final concentration of 7 μM.

Optical and EPR Spectroscopy.

The optical absorption spectra of membrane chromatophores were performed on a Shimadzu UV-2450 spectrophotometer. The samples were suspended in MOPS buffer and measured in the wavelength range of 500 to 600 nm. The ferricyanide was added to oxidize the samples, whereas for reduction, the sodium ascorbate and sodium dithionite were used (to reduce high- and low-potential b-type hemes, respectively). The spectra were recorded right after oxidation and after each step of reduction with ascorbate and dithionite at room temperature. CW EPR spectra of hemes were measured at X band frequency (9.4 GHz) using membrane chromatophores. Samples of WT and mutant were prepared in MOPS buffer. Samples were quickly frozen in liquid nitrogen and measured at 20 K. All EPR measurements were carried on a Bruker Elexsys E580 spectrometer using a SHQEOS11 resonator combined with ESR900 Oxford Instruments cryostat unit and the following parameters: resonance frequency, 9.39 GHz; microwave power, 1.9 mW; and modulation amplitude, 15 G.

Redox Potentiometry.

Redox midpoint potentials of hemes b were determined by dark equilibrium redox titrations on chromatophores according to the method described previously (39). Chromatophores were suspended in MOPS buffer and sustained under anaerobic conditions in the presence of redox mediators: 100 μM tetrachlorohydroquinone, 100 μM 2,3,5,6-tetramethyl phenylenediamine, 100 μM 1,2-naphtoquinone-4-sulfonate, 100 μM 1,2-naphtoquinone, 50 μM phenazine methosulfate, 50 μM phenazine ethosulfate, 100 μM duroquinone, 30 μM indigotrisulfonate, 100 μM 2-hydroxy-1,4-napthoquinone, 100 μM anthroquinone-2-sulfonate, and 100 μM benzyl viologen. Dithionite and ferricyanide were used to adjust ambient redox potential. During the EPR redox titration at certain ambient potential (Eh), the small aliquots of the sample (∼200 μL) were transferred to argon-flushed EPR tubes, quickly frozen in liquid nitrogen, and measured using parameters described above. The Em7 values of hemes b were determined by fitting the amplitudes of appropriate EPR g transitions to the Nernst equation for a one-electron couple with estimated uncertainties of approximately ±15 mV.

Computational Methods.

To complement our experiments on cytochrome bc1 complex from photosynthetic bacterium R. capsulatus, we performed fully atomistic MD simulations on the 3.5 Å resolution structure of the same complex (PDB ID: 1ZRT) (25). The protein was placed in a POPC lipid bilayer using the CHARMM-GUI online tool (40). An oxidized quinone (Q10) molecule was modeled at both the Qi and Qo sites in each monomer, and the [2Fe-2S] cluster and heme c1 were modeled in the reduced state, while the heme bL and heme bH were kept oxidized. Simulations with antimycin occupying the Qi site were also performed using the same crystal structure. Both WT and mutant (G48S) systems were constructed and simulated. Since the studied mutant is associated with mitochondrial dysfunction, we also performed MD simulations on the high resolution (1.9 Å) structure of cytochrome bc1 from yeast S. cerevisiae (PDB ID: 3CX5) (26) modeled in hybrid lipid bilayer consisting of 50% POPC, 35% POPE, and 15% cardiolipin. In both bacterial and mitochondrial model systems, all amino acids were kept in their standard protonation states, that is, Lys, Arg, Glu, and Asp in their charged states and all His residues in neutral state with Nδ protonated. The systems were solvated with TIP3P water molecules and 0.1 M NaCl salt concentration, resulting in a final system size of ∼285,000 atoms for bacterial systems (setups B1 to B4 in Table 2; see also ) and ∼580,000 atoms for mitochondrial systems (setups M1 and M2 in Table 2). CHARMM forcefield (41–43) was used for protein, lipids, water, and ions together with the force field parameters of antimycin obtained from CHARMM-GUI (40). The parameters of Q10 and metal centers were obtained from refs. 44 and 45, respectively. The MD procedure consisted of an initial energy minimization with constraints on heavy atoms, followed by 500 ps NVT (canonical ensemble, constant number of particles N, volume V and temperature T) and 1 ns NPT (isothermal-isobaric ensemble, constant number of particles N, pressure P and temperature T) equilibration steps, with constraints being gradually reduced throughout the equilibration. All simulations were performed with GROMACS 2016.4 and 2020.3 software (46) at 310 K and 1 atm pressure, with the temperature and pressure being controlled by the Nose-Hoover thermostat (47, 48) and Parrinello-Rahman barostat (49, 50), respectively. The electrostatic interactions were computed using the PME method (51) as implemented, and the LINCS algorithm (52) was used to achieve a step-size of 2 fs. The total production simulation time was ∼6 μs, and Table 2 shows details of the systems and simulation time lengths.

DFT calculations were performed to study the redox behavior of heme bH in WT and mutant states. Snapshots (n = 3) from MD simulations of bacterial complex were used to construct the WT and mutant cluster models for DFT calculations. Models comprised heme bH, its histidine ligands, and surrounding residues, totaling ∼200 atoms (). The total charge of the system in reduced and oxidized state was −1 and 0, respectively, and spin states considered were S = 0 and 1/2, respectively. Similar WT and mutant cluster models were also constructed using the high-resolution structure of yeast cytochrome bc1. DFT-based geometry optimizations were performed with the BP86 functional (53, 54) and def2-SVP basis set on all atoms other than Fe, where the larger def2-TZVP (55) basis set was used. All protein backbone atoms were removed apart from the Cβ atoms, which were fixed during geometry optimizations (). Dispersion corrections were applied (56) with Becke-Johnson damping (57). Single point calculations were then performed on the optimized structures using the def2-TZVP (55) basis set on all atoms, and three different density functionals, B3LYP (54, 58, 59), TPSS, and TPSSh (60), were used. The COSMO model (61) was employed to describe a dielectric constant of 4, mimicking the protein environment. All DFT calculations were performed using TURBOMOLE 7.3 software (62).