Introduction

Many electron transfer (ET) reactions occurring in vivo implicate pairs of proteins in which at least one of them is free to diffuse, like the small electron-carrying proteins plastocyanin, ferredoxin or cytochrome c (Cc)1. Specific recognition and binding between proteins occur widely in biology. To reconcile an overall sustained and efficient ET process demands a high turnover rate. Therefore, the complex formed between redox proteins in the electron transport chain (ETC) must be transient, and balance specificity and binding strength2,3. The recent direct experimental observation of electrochemically gated, long-distance ET between cytochrome c1 (Cc1, subunit of the cytochrome bc1) and Cc through the aqueous solution suggested that such tradeoff is achieved without direct contact, by molding the ionic distribution and electric field between their redox-active sites4. Thus, these redox protein partners might conciliate high specificity with weak binding, to keep the high turnover rate required by their biological function5.

Cc is a mitochondrial hemeprotein that exerts multiple functions, and its location depends on the cell condition6,7,8,9,10. Under homeostasis, Cc localizes in the intermembrane mitochondrial space, transferring electrons from cytochrome bc1 complex (Cbc1, complex III) to cytochrome c oxidase (CcO, complex IV) in the ETC. Two binding sites for Cc on Cc1 and CcO have recently been reported, a functional site (proximal site) enabling the correct orientation of redox centers to optimize the ET route, and a non-productive one (the so-called distal site) which is implicated in the turnover of Cc molecules11,12. Cc also takes part in other redox reactions within mitochondria, including reactive oxygen species (ROS) scavenging, redox-coupled protein import via Erv1-Mia40 pathway, and reduction of p66Shc generating ROS8. Upon distinct stimuli, such as oxidative stress or DNA damage, Cc can relocate to several cell organelles, interacting with proteins governing cell life and death13,14,15,16,17,18,19,20. Regarding oxidative stress, around 0.2-2% of the electrons involved in the ETC yield ROS via complex I (mainly), II, and III21. ROS have a dual role in cell signaling showing regulatory behavior at low levels and inducing apoptosis through oxidative stress at high ROS concentration. When mitochondrial production of hydrogen peroxide increases, Cc functions as a proapoptotic protein via its pseudo-peroxidase activity22. Around 15% of the Cc population is anchored to the inner mitochondrial membrane through binding to the lipid cardiolipin23,24. The pseudo-peroxidase function of the cardiolipin-Cc complex is activated at high levels of hydrogen peroxide and leads to the release of the hemeprotein from the mitochondria into the cytosol where it triggers apoptosis25,26. In mammals, Cc interacts in the cytosol with the apoptosis protease activating factor-1 protein (Apaf-1), to form the apoptosome and leading to downstream caspase activation and cell death8,27,28.

These moonlighting Cc activities are tightly regulated by posttranslational modifications (PTM)29,30. Notably, phosphorylation occurs in physiological and disease conditions and is highly tissue specific. Up to five tissue-specific phosphorylation sites of Cc have been characterized in mammals, including Thr28, Ser47, Tyr48, Thr58 and Tyr9731. Phosphorylation of Cc has been described to exert a cytoprotective and antiapoptotic role by downregulating ETC flux, thereby preventing mitochondrial membrane potential hyperpolarization, decreasing ROS production, and avoiding caspase activation. Phosphorylation of Cc has also been related to several pathologies, such as ischemia/reperfusion injury, neurological disorders, and cancer progression32. Although phosphorylation mainly occurs on serine and threonine residues with high frequency, phosphorylation of tyrosine residues has a crucial role in the regulation of mechanisms directly relevant to mitochondrial and cancer signaling, such as metabolic homeostasis, differentiation, and proliferation33,34. Among the two Cc tyrosine residues that can be phosphorylated in vivo, Tyr48 residue is located at the Ω-loop in the vicinity of the heme cleft, whereas Tyr97 lays far from the heme cleft in the C-terminal α helix of Cc. Cc phosphorylated on residue Tyr48 was primarily isolated from bovine liver tissue and it has been related to cancer35. Moreover, it has been found that Tyr48 can be mutated by histidine, which causes a mild human disease—thrombocytopenia—that results in a lower level of blood platelets36. Therefore, discerning the molecular mechanisms that regulate the function of phosphorylated Cc at position 48 is of great interest for understanding the role of Cc in pathology.

Functional studies revealed that Tyr48-phosphorylated Cc preserves the heme environment (no changes are observed in the Fe-Met80 absorption band at 695 nm) and that both Tyr48-phosphorylated and nonphosphorylated Cc species produced a hyperbolic kinetic response and the maximal turnover number of non-phosphorylated Cc was two times higher than that of phosphorylated Cc (3.7 s−1 and 8.2 s−1, respectively)35. The technical difficulties to preserve the in vivo protein phosphorylation after purification, the large amount of phosphoprotein needed for structural and biophysical assays and the lack of information on specific Cc-Tyr48 kinases requires adopting other approaches to study this phosphorylated protein. Over the last years, some authors have developed engineered aminoacyl-tRNA synthetase systems that can incorporate O-phosphotyrosine. However, this approach presents multiple challenges as phosphotyrosine poorly penetrates the cells due to its charge, it is unstable inside the cells and the incorporation efficiency by the aminoacyl-tRNA synthase is relatively low, producing a low yield of labelling (<5%) that is incompatible with single molecules techniques37. A convenient approach is using phosphomimetic mutants for functional studies38. The Cc Y48E mutant, in which Glu mimics the negative charge of the phosphate group, showed similar reaction kinetics to the non-phosphorylated Cc but presented a −45 mV midpoint redox potential shift. This mutation also inhibits the activation of caspase downstream signaling in vitro39. Nevertheless, the Y48E mutation leads to ca. 20% decrease in the volume of residue Tyr48, whereas it is increased upon phosphorylation. Thus, using Y48E to mimic phosphorylated Y48 underrepresents volume-related effects40. Therefore, the incorporation of a noncanonical amino acid that presents volume and charge values closer to those of phosphotyrosine is a better alternative. Among the different noncanonical amino acids that can be used to mimic Tyr-phosphorylation, p-carboxy-methyl-L-phenylalanine (pCMF) is one of the best substitutes, as it is stable, efficiently incorporated by E. coli cells, and preserves a similar volume than the phosphotyrosine residue41. This mutation does not affect the overall structure and heme iron environment as Tyr48-phosphorylated Cc (Supplementary Fig. 1)42. In terms of functionality, the Y48pCMF Cc shows lower binding affinity for the distal sites of Cc1 and CcO, altering the diffusion pathway of Cc molecules through Cbc1-CcO, thus modulating the ETC flux42. Finally, the Y48pCMF mutation effectively enhances peroxidase activity and induces an antiapoptotic function of Cc42.

Our recent observation of inter-protein long-distance ET between Cc1 and Cc through the aqueous solution4 can be explained by the presence of a charge conduit at the active interface between the two proteins (Gouy-Chapman conduit). These results suggest that ET could occur already when the two proteins are approaching (i.e., in the encounter state) without the need of establishing a well-defined, static protein complex, thereby reconciling high specificity with weak binding, and allowing to keep the high turnover rate required in the ETC5. Within this scenario, the presence of a charge conduit is of paramount importance for a sustained electron transport. The spatial extent of the charge conduit was found to depend on the redox potential of the interacting proteins and on the charge distribution between them.

Here we aim to investigate the molecular mechanism of ET regulation by phosphorylation in Cc-Cc1. We deploy an array of single-molecule, bulk, and computational methods to address this question in wildtype, phosphomimetic, and phosphorylated Cc. Our findings show that phosphorylation strengthens the interaction with Cc1 (lower equilibrium dissociation constant—KD—) and disrupts the Gouy-Chapman conduit associated with long-distance charge transport. These results could be of general significance for inter-protein ET, its regulation under diverse conditions in the cell, and the switching between multiple tasks.

Results

Long-distance ET between Cc-Cc 1 partners is impaired by Cc Tyr48 phosphorylation

The effects of Tyr48 replacement by the phosphomimetic pCMF in the inter-protein ET between Cc and the mitochondrial complex III, or Cbc1, were studied using the cross complex between human Cc and the soluble domain of plant cytochrome Cc1—which contains all residues of the physiological mature (without mitochondrial transit peptide) protein until Glu265, at which the C-terminal hydrophobic helix starts12—by means of electrochemical scanning tunneling microscopy (EC-STM)4. Cc and Cc1 were immobilized on the probe and the sample electrodes of EC-STM, respectively, with their redox active sites facing the solution, thereby allowing protein interaction. Cc1 was immobilized through its native N-terminal Cys10 to the Au(111) sample electrode of the EC-STM, and Cc or Y48pCMF Cc were anchored to the gold probe of the EC-STM through the C-terminal mutation E104C, thus facilitating the homogenous orientation of the proteins, while still preserving rotational freedom and good conductivity43.

The experiments were conducted as described4 in an electrochemical cell under bipotentiostatic control of the probe and sample electrodes versus an Ag/AgCl (SSC) reference electrode, spanning a potential range around the redox potentials of Cc1 and Cc, and always keeping Cc1 reduced (e.g. sample potential (US) = −0.20 V) and Cc oxidized (e.g. probe potential (UP) = 0.60 V). These values yield a constant positive bias Ubias = UP - US = 0.80 V by which the electrons are transferred from Cc1 to Cc, in analogy to physiological ET.

Current-distance (I-z) electrochemical spectroscopy was conducted between Cc1-Cc and Cc1-Y48pCMF Cc in 50 mM phosphate buffer pH 6.5. By positioning the EC-STM probe over the sample at a current set point of 0.4 nA, and after a stabilization period, the feedback loop was briefly disconnected, and the probe current recorded while the piezoelectric scanner holding the probe was retracted 15 nm from the set point at 12 nm·s−1. Then, the feedback was restored, and the probe reengaged, allowing subsequent recordings at different areas of the sample. Faradaic leakage current was maintained below a few pA over the entire recording range through probe insulation44. A nearly exponential current decay was observed in both cases from the initial current set point up to the distance-independent faradaic leakage current. For Cc1-Cc the current decay spans several nanometers as previously reported4, while for Cc1-Y48pCMF Cc the current decays more abruptly, reaching leakage values 3 nm away from the set point (Fig. 1a). Fitting the exponential region of the curves yields the distance decay factor (β) distributions shown in Fig. 1b, which are centered at β = 1.3 ± 0.8 nm−1 for Cc1-Cc, and at β1 = 1.3 ± 0.2 nm−1 and β2 = 3.1 ± 0.4 nm−1 for Cc1-Y48pCMF Cc showing a bimodal distribution (Fig. 1c). Although β1 values indicate that long-distance ET still takes place for a small population of recordings (up to 22%) in Cc1-Y48pCMF Cc, β2 population doubling the β obtained for Cc suggests that the replacement of Tyr48 by the phosphomimetic pCMF causes a severe alteration of the inter-protein ET. The presence of the phosphomimetic group can affect inter-protein ET by perturbing the two main steps involved in the process. Namely, modifying intra-protein conduction pathways at the protein level or altering the interface between the two proteins and thus affecting the charge conduit. When we compared the β values between bare gold and either Cc or Y48pCMF Cc, we found no significant differences (Supplementary Fig. 2), thus indicating that the Y48pCMF mutation does not modify protein conduction, and that the changes observed in Fig. 1a, b may involve the charge conduit that is established between Cc1 and Cc.

Fig. 1: Current-distance electrochemical tunneling spectroscopy of Cc1-Cc.
figure 1

a Ensemble of semi-logarithmic current-distance (I-z) curves during probe retraction, showing a more abrupt current decay for Cc1-Y48pCMF Cc. Sample and probe electrodes are represented by a square and a triangle, respectively. Us = −200 mV vs. SSC and at constant bias of 800 mV. b Histograms of distance decay factors (β) quantified from individual curves in a. c plot of the averaged β values (mean ± s.d.) β = 1.3 ± 0.8 nm−1 for Cc1-Cc (n = 103 I-z curves from 2 independent experiments, red) and β = 3.1 ± 0.4 nm−1 Cc1-Y48pCMF Cc (n = 64 I-z curves from 2 independent experiments, blue), respectively, showing significant differences (two-sample t-test, t Statistic = 19.65, degrees of freedom = 195, **** P < 0.0001). d Averaged β values (mean ± s.d. of n = 80) vs. EC gate potential at 200 mV constant bias obtained for Cc1-Cc (red) and Cc1-Y48pCMF Cc (blue). The midpoint redox potentials for Cc1 at the sample (0.276 V; green dashed line) and Cc or Y48pCMF Cc (red and blue dashed lines, respectively) at the probe are indicated, showing the β minimum for Cc1-Cc is located within the midpoint redox potentials of Cc1 and Cc. Despite being redox-active at similar potential than Cc, Y48pCMF Cc did not display a β minimum within the measurement range. All experiments were performed in 50 mM phosphate buffer, pH 6.5. Initial current set point 0.4 nA. Averaged β values for Cc1-Cc (red) are reproduced from reference 44. Source data are provided as a Source Data file.

Full size image

Electrochemical gating experiments were conducted for the Cc1-Y48pCMF Cc interaction. We recorded I-z plots at different probe and sample potentials keeping the bias constant. The potential of the reference electrode was adjusted to modulate the gate voltage in analogy to the gate electrode in a field effect transistor. As the EC-STM probe is grounded in our STM electronic configuration, the EC gate corresponds to -US at any given bias45. For Cc1-Cc we have previously observed that β displays a marked potential dependence at a moderate bias of 200 mV. Averaged β values showed a minimum (β = 0.5 ± 0.3 nm−1) at an EC gate potential of −0.25 V, which was located between the midpoint redox potentials of Cc1 (0.28 V vs. SSC) and Cc (0.35 V vs. SSC), respectively (Fig. 1d, red curve). This potential-dependent effect allowed to measure current spanning more than 10 nm away from the departure set point and highlights the contribution of the redox sites in the ET between Cc1 and Cc4. The β values of Cc1-Y48pCMF Cc do not depend on the EC gate within the considered redox potential window (Fig. 1d, blue curve), in striking contrast to Cc1-Cc. The midpoint redox potential of immobilized Y48pCMF Cc (0.36 V vs. SSC), is very similar to that of Cc (Supplementary Fig. 2), as previously reported41, and contrary to the Y48E substitution, which shows a ∼ 80 mV reduction of its midpoint redox potential with respect to Cc39. The redox potential values obtained are within the previously reported values for electron-transfer active Cc directly immobilized on an Au electrode46,47. Therefore, Y48pCMF Cc maintains its redox activity but cannot be engaged in long-distance ET through the aqueous solution. This is likely due to an alteration in the charge conduit between the protein partners rather than coarse protein structural changes. Cc structure consists of five cooperative folding-unfolding units (foldons)48 and its midpoint redox potential is very sensitive to protein folding and heme group coordination49,50. As Met80 is located in the second most unstable foldon (foldon IV) of Cc, subtle alterations of Cc structure affect the Met80-Fe coordination bond and, therefore, the redox potential value. Indeed, several authors have demonstrated the important role of Met80-Fe coordination on Cc redox potential, as the loss of this bond shifts the redox potential towards negative values (−200 mV)51. Based on our voltammograms (Supplementary Fig. 2), we can rule out the hypothesis of partial or substantial denaturation of the protein after its immobilization on the Au(111) surface, since no voltammetric signal is observed at negative redox potential values.

To complement current-distance spectroscopy measurements, we studied current blinks in the static break-junction approach52,53 to measure the spontaneous interaction between protein partners (Fig. 2). By setting the current set point to 0.30 − 0.35 nA, and transiently disconnecting the feedback loop at a constant bias of 0.8 V, we recorded current vs. time (I-t) traces on Cc1-Cc and Cc1-Y48pCMF Cc interaction (Fig. 2a, red and blue, respectively). In the I-t curves, sudden jumps appear in the measured current (blinks), which are attributed to the spontaneous contact between the probe and the sample. The collected blink currents (Iblink), after subtracting the baseline current (Ibaseline) were set to a common time origin (initial blinking time = 0) and plotted with the blinking lifetime in 2D-histograms (Fig. 2b). These 2D-blinking maps show the conductance dispersion of the partners’ interaction. Taking the most probable conductance peak, we found that Cc1-Y48pCMF Cc interaction led to a significantly higher conductance (P < 0.05) (Fig. 2c). Gaussian fitting of main conductance peaks yields G = (4.3·10−6 ± 9·10−7)·G0 for Cc1-Cc and G = (7.4·10−6 ± 5·10−7)·G0 for Cc1-Y48pCMF Cc (with G0 = 2e2/h). In Fig. 2d, Iblink is plotted against its corresponding Ibaseline showing that most probable Cc1-Cc interaction (G = 4.3·10−6 ± 9·10−7)·G0 can occur below the set point current (dashed line), thereby at longer distances than with Y48pCMF Cc, in agreement with I-z measurements.

Fig. 2: Blinking experiments.
figure 2

a Representative blinking recordings for Cc1-Cc (red) and Cc1-Y48pCMF Cc (blue). Blinks are shown as transient jumps of the current flowing between the two electrodes at a constant distance (imposed by initial current set point) and bias (800 mV). b Two-dimensional histograms of blink current (Iblink) and blink lifetime for Cc1-Cc (left, n = 100980 conductance values included) and for Cc1-Y48pCMF Cc (right, n = 252193 conductance values included). Counts have been normalized to a color scale, with 100 counts representing the maximum, and 0 representing the minimum. The corresponding conductance (G) profiles are shown. Baseline correction was applied in a and b. G = Iblink/Vbias is used to obtain the conductance values. G0 = 2e2/h = 77.5 μS, with h the Planck constant and e the elementary charge. c Plot of the averaged G values from gaussian fit of the most probable conductance peaks (mean ± s.d.) G = 4.3·10−6 ± 9·10−7 G0 for Cc1-Cc (red) and G = 7.4·10−6 ± 5·10−7 G0 for Cc1-Y48pCMF Cc (blue), respectively, showing significant differences (n = 6 conductance values from 2 independent experiments, two sample t-test, t Statistic = 2.777, degrees of freedom = 10.00, *P = 0.0195. d Plot of Iblink against its baseline current (Ibaseline) for a set of blinks of Cc1-Cc (red) and Cc1-Y48pCMF Cc (blue) (mean ± s.d.). The dashed vertical lines indicate the corresponding current set points at which feedback was turned off. Sample and probe electrodes are represented by a square and a triangle, respectively. All experiments were performed in 50 mM phosphate buffer, pH 6.5. Source data are provided as a Source Data file.

Full size image

Cc Tyr48 phosphorylation strengthens the Cc-Cc 1 interaction

Since EC-STM results suggest that Cc phosphorylation alters the interface between the two proteins, we further evaluated the inter-protein interaction using bulk and single molecule measurements. We assessed the effect of Tyr48 replacement by the phosphomimetic pCMF on the interaction strength between Cc and Cc1 by analyzing the forced unbinding between human Cc and the soluble domain of plant cytochrome Cc1 using single molecule force spectroscopy with an atomic force microscope (AFM)54. In addition, binding affinities of the Cc1-Cc complex were calculated from solution-based surface plasmon resonance (SPR) ensemble measurements, and the impact on Cbc1 complex (complex III) activity was assessed in mammalian mitochondrial extracts.

To carry out single-molecule unbinding experiments, Cc1 was immobilized on Au(111) substrates in a similar way as described for the EC-STM experiments. Either Cc or Y48pCMF Cc, both with the C-terminal mutation E104C, were conjugated to the maleimide group of a PEG27 heterobifunctional crosslinker previously anchored to amino-functionalized AFM silicon probes. In the force spectroscopy measurements, the tip and substrate are approached and forced into contact up to a maximum force of 200 pN and then retracted away from the surface at constant velocity, in 50 mM phosphate buffer pH 6.5. The force-separation retraction traces at 1 μm·s−1 velocity occasionally displayed unbinding events like those in Fig. 3a, which allows quantifying the force (Fr) and length (lr) (see representative example in Fig. 3a inset) at which the contact is ruptured. The fact that the rupture length is maintained around 10–20 nm (Fig. 3b) indicates that protein unfolding is not significant, considering that the linker length is about 8 nm (monomer length of approx. 0.3 nm)55,56, and the measured distances may be due to deformation and elongation of the proteins prior to dissociation. No correlation is found between Fr and lr for neither Cc1-Cc nor Cc1-Y48pCMF Cc (Fig. 3b). For the pair Cc1-Cc, Fr is centered at 55 ± 17 pN (fitting a gaussian distribution) with a mean value of 75 pN (Fig. 3b, c). These relatively low rupture forces support the idea of the transient nature of the Cc1-Cc interaction. In contrast, Cc1-Y48pCMF Cc display higher unbinding forces (P < 0.05), centered at 77 ± 24 pN (gaussian fit) with a mean value of 91 pN (Fig. 3b, c).

Fig. 3: Unbinding between Cc1-Cc studied by AFM-FS single molecule and surface plasmon resonance (SPR) ensemble measurements.
figure 3

a Representative force-separation unbinding events upon tip retraction at 1 μm·s−1 for Cc1-Cc (red) and Cc1-Y48pCMF Cc (blue), when Cc1 is directly anchored to Au(111) and Cc is anchored to the AFM tip through a PEG27 linker. The inset shows the parameters Fr and lr, that are collected from each curve and displayed in b as Fr vs. lr with the corresponding histograms in each axis. Fr distributions are fitted to a Gaussian. c Plot of the averaged Fr values (mean ± s.d.) for Cc1-Cc (n = 134 force-separation curves from 2 independent experiments, red) and Cc1-Y48pCMF Cc (n = 88 force-separation curves from 3 independent experiments, blue), showing significant differences (two sample t-test, t Statistic = 2.34, degrees of freedom = 220, *P = 0.02). d Fr vs. lr contour maps for the unbinding at different pulling rates (0.5 − 4.0 μm·s−1) for Cc1-Cc (top, n = 212, 185, 55 and 226 for 0.5, 1.0, 2.0 and 4.0 μm·s−1, respectively) and Cc1-Y48pCMF Cc (bottom, n = 194, 83, 72 and 97 for 0.5, 1.0, 2.0 and 4.0 μm·s−1, respectively), when Cc1 is directly anchored to Au(111) and Cc is directly anchored to the Au-coated AFM tip. e Mean Fr vs. tip velocity (in log scale). f Average Fr (mean ± s.d.) vs. ln (loading rate) for the Cc1-Y48pCMF Cc unbinding fitted to a linear function (blue line) (n = 194, 83, 72 and 97, for 28000, 56000, 112000 and 224000 pΝ·s−1, respectively). Forced-activated unbinding parameters are calculated following the Bell-Evans model57,59,62. All AFM experiments were performed in 50 mM PBS, pH 6.5. g Binding kinetics determined from SPR measurements. SPR sensograms obtained with 5 µM (squares) and 2.5 µM (circles) of Cc (red) and Y48pCMF Cc (blue) analyte and immobilized Cc1. Lines correspond to a simultaneous, global fit of the two curves to a one-site binin model. Data is representative of a total of 16 spots. The binding assays were performed in 10 mM PBS, pH 6.5. Source data are provided as a Source Data file.

Full size image

We complemented these results with force-mediated dissociation kinetics studies. The lifetime of a bond sustained by weak noncovalent interactions decreases when subjected to force, which is conceptualized as force causing a tilt on the energy landscape, lowering energy barriers57. The unbinding force generally depends on the loading rate, as theory predicts for many biological interactions like receptor-ligands bonds58. In that case, the measured unbinding force corresponds to an individual point in a continuous spectrum of bond strengths. Therefore, we measured the rupture force at varying retraction velocities (from 0.5 to 4 μm·s−1). To avoid transmission of the load through the polymer linker and bring high inaccuracy to the loading rate values, either Cc or Y48pCMF Cc were immobilized directly onto Au-coated probes through the thiol-Au chemistry. Here, the protein immobilization and orientation are comparable to those of the ECSTM experiments. The two proteins are then forced into contact and Fr and lr are evaluated at different retraction velocities. These measurements only carry direct information of the kinetics of unbinding (not binding).

The force of Cc1-Cc unbinding is constant for the pulling velocities studied (contour maps Fr vs. lr, Fig. 3d). In contrast, Cc1-Y48pCMF Cc unbinding becomes force resistant, and the unbinding force scales logarithmically with the pulling velocity and the loading rate (Fig. 3d). Cc1-Cc dissociation occurs at low forces independently of the pulling velocity (mean Fr vs velocity in log scale, Fig. 3e). This behavior is associated with two bodies being pulled apart during a near-equilibrium mode59,60,61, in which complexes may have short lifetime and unbinding-rebinding is faster than pulling. This allows quick dissociation of the complex at biologically accessible loading rates and forces, in accordance with the requirement of a rapid turnover. In contrast, the rupture force between Cc1-Y48pCMF Cc depends logarithmically on the loading rate (r) (Fig. 3f), meaning that the measurements are done in a kinetic regime (non-equilibrium bond rupture), as predicted by the Bell-Evans model based on a single transition state theory57,59,60,62. The most probable unbinding force plotted against the loading rate on a log scale results in a straight line. Kinetic parameters can be obtained assuming that the energy barrier observed in dynamic force spectroscopy (DFS) remains rate limiting below the loading rate range tested. Following the Bell-Evans model57,59,62, the rate of the molecular complex force-activated unbinding (kou) and the distance from the bound state to the barrier that is rate limitng for unbinding (xu) can be calculated from a linear fit to the data only for the Cc1-Y48pCMF Cc pair (kou = 48 s;−1 xu = 0.31 nm). Overall, the DFS results indicate that the Y48pCMF mutation in Cc strengthens the interaction with Cc1, thereby hindering a fast turnover.

Cc1-Cc binding affinities were calculated from solution-based surface plasmon resonance (SPR) ensemble measurements. Protein binding affinities are commonly measured and described by the equilibrium dissociation constant (KD), which evaluates the strength of biomolecular interactions (i.e., the greater the KD value, the weaker the attraction between both protein partners). Fitting of the SPR curves for the interaction between Cc (WT and Y48pCMF) and Cc1 to a 1:1 binding model (Fig. 3g) allows to obtain the association (kon) and dissociation (koff) rates for the interaction (Table 1). The calculated association rate kon of Cc1-Y48pCMF Cc complex is nearly 2-fold larger than that of the Cc1-Cc complex, whereas the differences between spontaneous off-rate koff values are small (but statistically significant, P < 0.05). Differences found between kou values determined in single-molecule DFS experiments and spontaneous off-rates koff values determined by SPR have been attributed to the complexity of the energy landscape that may cause high differences between the forced unbinding mechanism and the thermally activated one63,64.

Table 1 Analysis of surface plasmon resonance (SPR) curves
Full size table

The differences in KD (koff/kon) values between Cc and Y48pCMF Cc in complex with Cc1 go along with the single molecule force spectroscopy results, and agree with published NMR and ITC measurements42, demonstrating a stronger interaction force between Cc1-Cc when the Tyr48 residue is replaced by pCMF.

We further investigated if the observed differences in the unbinding forces and KD values could modulate the mitochondrial complex III activity (Supplementary Fig. 4). The biochemical assay to study the complex III activity relies on following the changes in Cc redox state from oxidized to reduced, by recording the absorption spectrum of Cc at 550 nm. Our results with mammalian mitochondrial extracts showed that Y48pCMF Cc displays a 1.6-fold reduction (P < 0.001) of the net complex III activity (0.28 nmol·min−1 μg−1 vs. 0.17 nmol·min−1·μg−1 for WT and Y48pCMF Cc, respectively). Such results support the hypothesis that the observed differences in unbinding forces and KD values could be physiologically relevant in the modulation of the ET behavior of Tyr48-phosphorylated Cc.

Cc phosphorylation disrupts the Gouy-Chapman conduit for long-distance ET with Cc 1

Long-distance currents between Cc and Cc1 can be explained by a low ionic concentration in the aqueous gap between the two protein redox sites, compared to the solution bulk4. Cation depletion near the redox-active sites impairs charge screening in the solution and allows the electric field to extend several nanometers within the diffuse ionic Gouy-Chapman layer. To account for the changes in β observed when Tyr48 is replaced by the phosphomimetic pCMF in the EC-STM experiments, we performed molecular dynamics (MD) calculations of the pairs Cc1-Cc (WT) and Cc1-Y48pCMF Cc (phosphomimetic), as well as Cc1-pY48 Cc (phosphorylated Tyr48: O-phospho-L-tyrosine) with charge −1 and −2. Note that the latter “true” phosphorylated Cc forms offer valuable insights as they are not yet experimentally accessible. The proteins were separated 3.2 nm with the redox active sites facing each other, in a solvent with explicit water molecules and 50 mM NaCl concentration (Supplementary Fig. 3a–d)

Cation depletion near the redox-active sites in the aqueous gap between the proteins is observed for Cc1-Cc, and the concentration is nearly zero in the proximity of the redox active sites (Fig. 4a), in accordance with our previous studies4. In contrast, for Cc1-Y48pCMF Cc an accumulation of cations is found around the proteins and in the region between them (Fig. 4b, h), likely due to the extra negative charge at position 48 that alters the surface electrostatic potential. This is reflected in the computed electrostatic potential in a 50 mM ionic concentration, which shows equipotential lines that extend between Cc and Cc1 proteins, forming a conduit (Fig. 4c), but this is reduced between Y48pCMF Cc and Cc1 (Fig. 4d). For Cc1-Cc the space between both heme groups had the lowest average concentration of sodium ions (0.018 M) in comparison to Cc1-Y48pCMF Cc (0.067 M). Thus, MD calculations show that the Y48pCMF mutation of Cc causes the disruption of the Gouy-Chapman conduit that has been associated to long-distance ET through the aqueous solution in agreement with EC-STM results. Interestingly, the phosphorylation of Tyr48 in native Cc may shift the electrostatic field even more than the Y48pCMF mutation, as evidenced when the computed average sodium concentration (average sodium concentration of 0.044 M for charge −1 and 0.052 M for charge −2, Fig. 4h, Supplementary Fig. 3e, f) and electrostatic potentials are obtained for pY48 Cc with charge −1 and −2 (Figs. 4e, f, h). In both cases, the equipotential lines do not extend between Cc and Cc1.

Fig. 4: Molecular dynamics (MD) calculations of Cc1 and Cc display an inter-protein Gouy-Chapman conduit that is disrupted by phosphorylation.
figure 4

a, b Side view of the averaged sodium concentration map (M) at the X-axis plane dissecting both Cc1-Cc redox active sites (Cc and Y48pCMF Cc, respectively) facing each other at 3.2 nm separation taking the most external hydrogen atoms from the -CH3 distal pairs from each heme group, that is 4.2 nm taking the Fe-to-Fe distance. MD simulations were performed with explicit water (TIP3P solvent model) and 50 mM ionic concentration (NaCl) over 36 ns. Each contour line displays an increase of 0.15 M. (c–f) Side view of computed electrostatic potential with APBS76 from –1 kT/e (red) to 1 kT/e (blue) for Cc1-Cc pairs: Cc1-Cc (c), Cc1-Y48pCMF Cc in silico (d), Cc1-pY48 Cc (O-phospho-L-tyrosine) with charge –1 (e) and Cc1-pY48 Cc (O-phospho-L-tyrosine) with charge –2 (f). Ion concentration is 50 mM, pH 6.5. Each contour line displays an increase of 0.2 kT/e. g Virtualization of the Cc1-Cc system in 3D. Representation of the prism (blue) used to calculate the average concentration of sodium ions throughout the simulation. The image shows a single frame of the Cc1-Cc system depicting protein (cyan, translucent grey surface), heme groups (white with Iron atom in red), sodium ions (green spheres), and omitting water molecules. The closest backbone residues to the generated prism vertices are His18, Pro76, Lys79 and Ile85 of the Y48pCMF Cc, and Leu134, Leu94, Glu101 and Ala160 of Cc1. h Average sodium concentration over 36 ns in a 12.9 nm3 volume prism (prism dimensions for x, y, z = 15, 43, 20 units) centered between both heme groups of Cc and Cc1 for all systems. Volmap Tool Plugin 1.1 from VMD75 was used. i Derivation of binding kinetics from the BD computations of Cc1-Cc (red) and Cc1-Y48pCMF Cc (blue) complexes. For each computation, two curves are represented, corresponding to 4 independent pair-distances criteria for reaction, as reported previously65. Vertical lines correspond to the distances used as reaction criterion to estimate kon and ktrigger values. Source data are provided as a Source Data file.

Full size image

We also performed Brownian dynamics (BD) simulations to calculate binding kinetics using the MD structures of Cc and Y48pCMF Cc complexes with Cc1. For each computation, two curves are represented (Fig. 4i), corresponding to 4 independent pair-distances criteria for reaction, as reported previously65. We estimated association rate (kon) values at a distance of 0.6 nm as reaction criterion (vertical line in Fig. 4i). Cc and Y48pCMF Cc species show different behaviors across the simulation volume (expressed as distance cut-off dependencies of observables): i) the long-range attraction between Y48pCMF Cc and Cc1 is larger than between Cc and Cc1, and ii) below 12 nm, the estimated apparent kon values for Y48pCMF Cc decay more steeply than in the case of Cc, probably due to bias of electrostatic steering. This matches the differences in force values found in DFS experiments, indicating that the Y48pCMF mutation strengthens the interaction with Cc1, as pointed out before. The ratio between the kon values decreases hyperbolically with the distance, reflecting the impact of changes in the energetic landscape of interaction. The rate constant ktrigger was set at the cut-off distance (2.5 nm) at which ET can be triggered in the EC-STM experiments. At the contact level, reckoned kon for Cc1-Y48pCMF Cc are ca. 10-fold larger than those estimated for Cc1-Cc (Table 2, Fig. 4i). At the cut-off distance of 2.5 nm, ktrigger values increase 2-3-fold in the trajectories corresponding to Cc1-Y48pCMF Cc compared to Cc1-Cc. In summary, Cc1 attracts Y48pCMF Cc stronger than the WT Cc, even at long distances.

Table 2 Analysis of brownian dynamics (BD) simulations for the Cc1-Cc interaction
Full size table

Discussion

Phosphorylation is one of the most prevalent mechanisms of regulation of protein activity. In particular, tyrosine phosphorylation plays a crucial role in cellular processes such as metabolism or cellular growth. Cc and its partner CcO are among the mitochondrial proteins whose function is highly regulated by phosphorylation. Cc phosphorylation has been suggested to perform a crucial role in regulating mitochondrial respiration and cell fate30.

Tyr48 phosphorylation has been reported to affect the binding affinity of Cc towards its molecular partners, altering the diffusion pathway of Cc molecules through Cbc1-CcO and affecting the ETC flux by partially inhibiting the mitochondrial respiration42. However, the molecular mechanism through which Cc phosphorylation impacts the ETC is unknown. The multimodal experiments on phosphomimetic Cc and calculations reported here show that upon phosphorylation the long-distance charge conduit established between Cc-Cc1 is shut off and their interaction is strengthened and departed from binding/unbinding equilibrium in the measured conditions. Overall, these results are in full agreement with the consensus evidence published on the topic and unveil a nanoscopic view and mechanistic framework that offers deeper insights into the interaction between redox protein partners of the respiratory ETC.

Notably, these results indicate a regulatory advantage of using electrochemically gated, long-distance electron transport through the solution between redox partner proteins. Besides reconciling high specificity and electron transport efficacy with weak binding to keep high turnover rates4, the Gouy-Chapman conduit bearing reduced ionic density in the confined volume between the proteins can be externally controlled to reduce charge flow. When phosphorylated at Tyr48 (herein mimicked by replacement with pCMF), Cc sliding from Cbc1 to CcO is impaired42. The stronger interaction between the partners upon phosphorylation was considered to hinder the Cc turnover (exchange with the partner) as it would lead to a frozen complex as concluded from calorimetry data and NMR spectroscopy results42. Note that our scanning probe microscopies experiments allow probing interactions at the single protein level, and in the case of EC-STM (I-z and blinking studies) it is possible to keep Cc1 reduced, and Cc oxidized to mimic the physiological situation of ET from Cc1 to Cc. Independent control of probe and sample electrochemical potential in the single molecule force spectroscopy AFM setup is more challenging66,67 and constitutes one aspect to improve experimentally in future studies.

We observe that electrochemically gated long-range ET is impaired in single-molecule experiments with phosphomimetic Y48pCMF Cc, with current decay distances reducing to half (to 3 nm or less, Fig. 1a–c) and abolishing EC gate potential control (Fig. 1d). MD simulations show that the negative charge introduced by the pCMF residue on the Cc surface prevents cation exclusion between the proteins’ redox active sites, thereby enabling charge screening in this confined interface and disrupting the Gouy-Chapman conduit (Fig. 4). The effects are even stronger in simulations with phosphorylated residues, which bear high relevance as they are not yet experimentally accessible. Besides, we observe that the single molecule unbinding of Y48pCMF Cc from the redox partner Cc1 occurs at higher forces (Fig. 3). These results are consistent with the higher binding affinities of phosphomimetic Cc compared to the wildtype in bulk experiments (Fig. 3g, Table 1) and BD calculations (Fig. 4i, Table 2).

Figure 5 recapitulates all these results and provides a mechanistic framework showing how phosphorylation of Cc alters several functions of this multifaceted protein involved in electron transport and cell fate. First, by shutting down its ability to exchange charge with Cc1 (complex III) at long distances through the aqueous solution and in an electrochemically selective manner4. And second, by promoting a stronger contact with Cc1 that disturbs the reversible binding found in equilibrium conditions (low forces) and likely reduces the turnover of molecules at complex III. In this scenario, even small amounts of phosphorylated Cc will lead to one of them being stuck at complex III and effectively preventing others from transferring charge. The dissociation constant (KD) is lower for the Cc Y48pCMF variant, which must get closer to complex III than the wildtype in order to achieve similar current levels (due to the higher β of the former, see Fig. 1a). In this situation, the interaction forces (unbinding force, Fig. 3; binding affinity, Table 1 and Fig. 3g; association rate, Table 2 and Fig. 4i) are higher and thereby the overall turnover of Cc molecules is reduced, lowering the complex III activity (Supplementary Fig. 4). The higher conductance observed in the Y48pCMF variant compared to the WT species (Fig. 2c) might be caused by such closer interaction.

Fig. 5: Proposed model showing how phosphorylation of cytochrome c promotes a closer contact with cytochrome c1 and diminishes the turnover of molecules at complex III.
figure 5

(Left) WT cytochrome c (red) can receive electrons from cytochrome c1 of complex III at long distances (low β). Overall turnover of cytochrome c molecules at complex III is high due to reversible binding in equilibrium conditions (low F). (Right) Compared to WT Cc (red), the dissociation constant (KD) of the Y48pCMF variant (blue) is lower and it must get closer to complex III for electron transfer (high β). In this situation, the interaction forces are higher (high F) and thereby the overall turnover of electron carriers is reduced. Conductance of the Y48pCMF variant is higher (right, thick black arrow), compared to that of the WT species (left, thin black arrow). Thus, phosphorylation of cytochrome c disrupts the long-distance electron transfer conduit (marked in yellow) previously observed in the wild-type protein4.

Full size image

In conclusion, we have performed a multimodal study of the interaction between redox partners Cc and Cc1 that covers single protein conductance and force measurements, bulk binding and biochemical assays, and molecular dynamics calculations. By studying a phosphomimetic mutation, we have focused on the impact of phosphorylation in all the observables and integrated the obtained results within the available biochemical evidence, aiming to provide a nanoscopic view and mechanistic framework that was lacking in this topic. Phosphorylation of Cc impairs electrochemically gated long-distance electron transport through the solution and disturbs the binding/unbinding equilibrium between the partners, strengthening their interaction. Calculations show that the cation exclusion volume that is established between the partners (Gouy-Chapman conduit) is disrupted by the negative charges introduced near the Cc redox site by phosphorylation and its experimental molecular mimics. These results allow a deeper understanding of the tradeoff between binding, transfer, and turnover that is established between redox protein partners in the respiratory ETC and provide mechanistic insights into the molecular sites that regulate cell signaling circuits.

Methods

Protein expression and purification

Escherichia coli (E. coli) BL21 (DE3) cells were transformed with pBTR1-WT plasmid—that contained the CYCS gene-coding for human Cc and, the CYC3 gene of the yeast Cc heme lyase, required for the maturation of Cc—to recombinantly express Homo sapiens WT Cc for Surface Plasmon Resonance (SPR) experiments. For the expression of E104C mutant, used for Electrochemical Scanning Tunneling Microscope (EC-STM) and Atomic Force Microscopy-Force Spectroscopy (AFM-FS) measurements, pBTR-1 plasmid was mutated by replacing the GAA triplet corresponding to Glu104 by TGT, which codifies to cysteine residue. The primers for PCR were pBTR1-E104C fw (5- AAGGCGACGAACTGTTGATAAGGTACCA-3) and pBTR1-E104C rv (5- TGGTACCTTATCAACAGTTCGTCGCCTT-3). For this purpose, one-step mutagenic PCR with Accusure™ DNA Polymerase (Bioline) was used following the manufacturer’s instructions. The new plasmid, containing the mutated sequence, is designated as pBTR1-E104C. Expression and purification of both Cc species was carried out as previously described68. Briefly, transformed bacteria were incubated for 20 h at 150 rpm and 30 °C in 2.5 L standard lysogeny broth (LB) rich medium (10 g·L−1 tryptone, 5 g·L−1 yeast extract and 10 g·L−1 NaCl), supplemented with 100 μg·mL−1 ampicillin. Cells were harvested via centrifugation at 5000 x g for 10 min at 4 °C, resuspended in 10 mM tricine buffer pH 8.5 supplemented with cOmplete protease inhibitor cocktail (#11836145001, Roche), ruptured by sonication for 5 min, and then centrifuged at 14,000 x g for 60 min. The supernatant was loaded onto a Nuvia S column (#7324720, Biorad) and eluted with a 25 to 500 mM NaCl gradient in 10 mM tricine buffer pH 8.5. Phosphomimetic Y48pCMF Cc was expressed using the evolved tRNA synthetase technique, as previously described41. Briefly, E. coli BL21 (DE3) cells were co-transformed with either pBRTR1-Y48AMBER or pBTR1-Y48AMBER/E104C, and pEVOL/pCMF/tRNA plasmids—which contains the unnatural tRNA/aminoacyl-tRNA synthetase pair that recognizes the pCMF—and were cultured for 20 h at 150 rpm and 30 °C in M9 minimal medium supplemented with 100 μg·mL−1 ampicillin, 20 μg·mL−1 chloramphenicol, 16.7 μg·mL−1 δ-aminolevulinic acid hydrochloride and 263 μg·mL−1 non-canonical amino acid pCMF. Protein expression was induced with 1 mM isopropyl-β-D-thiogalactoside and 0.02% arabinose. Cells were harvested by centrifugation and purified similarly as previously described for Cc41. The soluble domain of Arabidopsis thaliana Cc1 was expressed and purified as previously described12. Briefly, E. coli BL21 (DE3) cells were co-transformed with pET-Cc1 and pEC86 plasmids (which constitutively express the eight cytochrome c maturation genes ccmABCDEFGH) to produce the soluble domain of Cc1 (residues 64–265, Gene ID 834081). The cells were grown for 24 h at 150 rpm and 30 °C in 2 L of standard LB rich medium, supplemented with 50 μg·mL−1 kanamycin, 12 μg·mL−1 chloramphenicol and 16.7 μg·mL−1 δ-aminolevulinic acid hydrochloride. Harvested cells were resuspended in 20 mM Tris-HCl buffer pH 8.0 supplemented with cOmplete protease inhibitor cocktail (#11836145001, Roche) and physically ruptured by sonication for 5 min. The supernatant was loaded onto a Nuvia Q column (#7324721, Biorad) and eluted with a 25 to 500 mM NaCl gradient in 20 mM Tris-HCl buffer pH 8.0. Fractions containing Cc1 were further purified by size exclusion with a ENrich SEC 70 column (#7801070, Biorad). All protein samples were dialyzed against 10 mM sodium phosphate buffer at pH 6.5.

Electrochemical scanning tunneling microscopy-spectroscopy

Sample preparation

Atomically flat Au(111) single-crystal disks (10 mm diameter and 1 mm thickness, MaTecK) were flame annealed and electrochemically polished45. The Au(111) electrodes were incubated with Cc1 in sodium phosphate buffer 10 mM, pH = 6.5 at a 56 μM concentration. The EC-STM probes (Au wire 0.25 mm in diameter; GoodFellow) were insulated with ApiezonTM45, and were incubated with a 12.5 μM solution of either Cc (E104C) or Y48pCMF Cc (E104C), in sodium phosphate buffer 10 mM, pH = 6.5. All incubations were conducted overnight and at 4 °C.

EC-STM measurements

All experiments were performed with a PicoSPM microscope head and a PicoStat bipotentiostat (Agilent, USA) controlled by Dulcinea electronics (Nanotec Electronica, Spain) using the WSxM 4.13 software. A custom-made electrochemical liquid cell with a standard sample plate was used in four-electrode configuration with the STM Au(111) sample and probe as working electrodes, a 0.25 mm diameter Pt80/Ir20 wire as counter electrode and a miniaturized ultralow leakage membrane Ag/AgCl (SSC) reference electrode filled with 3 M KCl. The potentials of the Au(111) electrode sample (US) and EC-STM probe (UP) were expressed against this reference. The electrochemical cell was cleaned with piranha solution (3:1 v/v solution of H2SO4 and H2O2) before each measurement. Caution: piranha solution is a strong oxidizer and a strong acid. It should be handled with extreme care, as it reacts violently with most organic materials. The experiments were performed in sodium phosphate buffer 50 mM (pH = 6.5) electrolyte solution, previously filtered through 0.02 μm diameter sterile membrane (Anotop) and degassed with nitrogen.

The Current versus distance (I-z) curves were acquired by setting an initial current set point of 0.4 nA, turning the feedback loop off, and recording the probe current at 12 nm·s−1 during probe retraction along 15 nm (20 samples per point, 1024 points) at a constant bias (Ubias = UP - US). At least 2 independent experiments were conducted in each case. For each independent experiment, a total of 150 curves (20 samples per point, 1024 points) were recorded. The data were treated with the custom-generated MATLAB code (The MATHWORKS, Inc.) previously reported4.

For static blinking experiments, a NI-DAQmx and BNC-2110 LabVIEW setup was used for data acquisition, LabVIEW software was used for data treatment. The STM probe is set at a fixed current set point from the sample, and at a constant bias, and the feedback is turned off after a period of mechanical stabilization. Current versus time traces are then recorded in captures of few seconds during several hours. Current blinks are observed in the current transient in the form of telegraphic noise (Fig. 2a). At least 2 independent experiments were conducted in each case. The current is transformed to conductance values using G = Iblink/Ubias. Up to a hundred blinks are used to build 2D-blinking maps.

The data were analyzed with OriginPro 8.5.0 SR1 (OriginLab Corp.). n indicates the sample size in all cases. Errors are indicated as standard deviation of the mean.

Atomic force microscopy-force spectroscopy

Sample preparation

Cc1 was immobilized on atomically flat Au(111) single-crystal disks (10 mm diameter, 1 mm thickness, MaTecK). The Au(111) disks were electrochemically polished45 and flame annealed, cooled under a nitrogen flow and then incubated with a 56 μM solution of Cc1 in sodium phosphate buffer 10 mM, pH 6.5 within a humid environment, overnight and at 4 °C.

AFM probes preparation

We used V-shaped Si3N4 cantilevers (Bruker AFM Probes, Camarillo, CA) having a nominal spring constant of 0.06 N·m−1 with either silicon (SNL probes) or Au-coated (NPG probes) tips.

Au-coated probes were placed on a PDMS surface and incubated with a 25 μL drop of a 26 μM solution of Cc (E104C) or Y48pCMF Cc (E104C) in sodium phosphate buffer 10 mM, pH 6.5. Incubations were performed at 4 °C overnight in a humid environment.

Silicon probes were first treated for 10 minutes in a UV/Ozone ProCleanerTM (BioForce Nanosciences). Then, the probes were aminofunctionalized in atmosphere of 3-aminopropyl)triethoxysilane (APTES) and triethylamine (TEA). For this, the probes were placed inside the desiccator previously flooded with N2 gas to remove air and moisture. Two small plastic containers with 45 μL of APTES and 15 μL of TEA, separately, were placed close to the probes (on a clean inert surface) and vacuum was on for 5 min. After 1.5 h, the APTES and TEA container were removed, and vacuum was restored. The probes were left in the desiccator overnight at room temperature. The probes were then rinsed with chloroform (2x) and ethanol (2x) and dried under a gentle N2 flow. When the tips were not used immediately, they were stored in a desiccator under Ar atmosphere. The AFM probes were then placed in glass wells filled with 0.5 ml solution of 1 mg·mL−1 heterobifunctional PEG (n = 27) (Mal-PEG27-NHS, O-[N-(3-maleimidopropionyl)aminoethyl]-O′-[3-(N-succinimidyloxy)−3-oxopropyl]heptacosaethylene glycol, Sigma-Aldrich) and 5 μL TEA in chloroform and left for 2 h. The probes were then rinsed with chloroform (2x), ethanol (2x) and MilliQ water and dried under a gentle N2 flow. The probes were placed onto a PDMS surface and incubated with a 25 μL drop of a 26 μM solution of Cc (E104C) or Y48pCMF Cc (E104C) in sodium phosphate buffer 10 mM, pH 6.5. Incubations were performed at 4 °C overnight in a humid environment.

AFM-FS measurements

Force-separation curves were measured with either an MFP3D AFM (Asylum Research, Oxford Instruments) or a NanoWizard 3 BioScience AFM (JPK Instruments, Bruker Nano GmbH) at room temperature and under liquid environment (sodium phosphate buffer 50 mM, pH 6.5). After having measured the sensitivity (V·m−1), the cantilever spring constants were individually calibrated by using the equipartition theorem (thermal noise routine)69. Force-separation curves were recorded by approaching and retracting the AFM tip at constant velocity (0.5 to 4 μm·s−1) and in the force map mode over an area of 3 × 3 μm2. The maximum applied (contact) force in each cycle was set to 300 pN. At least two independent experiments were conducted in each case. AFM data were acquired and treated using the AFM software and analyzed with OriginPro 8.5.0 SR1 (OriginLab Corp.). Errors are indicated as the standard deviation of the mean.

Surface plasmon resonance SPRi

Cc interaction to Cc1 was assayed by SPR using a CS SPRi-Biochip™ (HORIBA Scientific) and a SPRi-Plex II™ system (HORIBA Scientific), as previously described with minor modifications70. Briefly, Cc1 direct amine immobilization onto the biochip was achieved by deposition of 10 and 15 µM solution drops on the surface and rabbit serum albumin (Sigma-Aldrich) was used as control at reference spots. The binding assays were performed at 298 K in 10 mM sodium phosphate buffer, pH 6.5. WT or Y48pCMF Cc were flowed at various concentrations (from 2.5 to 10 µM) over the modified surface with Cc1 at 50 µL·min−1 rate. The signals from the reference spot surface were subtracted in every sensogram. The data were analyzed and fitted to a one-to-one binding model adapted from Bowles et al71. with Origin 2018 (OriginLab Corporation).

Cytochrome bc1 complex activity

The ability of WT and Y48pCMF Cc variants to act as electron acceptor of cytochrome bc1 complex (complex III) was tested with mammalian mitochondria extracted from HeLa cells. HeLa cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Sigma-Aldrich) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Sigma-Aldrich), 2 mM L-glutamine (Sigma-Aldrich), 100 U·mL−1 streptomycin (Sigma-Aldrich) and 100 μg·mL−1 penicillin (Sigma-Aldrich). Mitochondria were isolated from HeLa cells as previously described and stored at −80 °C in 10 mM Tris-MOPS, 1 mM EGTA/Tris, 200 mM sucrose pH 7.472. Complex III activity was measured spectrophotometrically (Varioskan™ LUX microplate reader, Thermo Scientific™) using the commercial mitochondrial complex III activity kit from Abcam (ab287844), according to the manufacturer’s instructions. Net complex III activity was calculated by comparing the Cc reduction in the presence and absence of Antimycin A, a complex III specific inhibitor.

Molecular dynamics simulations

The initial model of the Cc and Cc1 proteins was done essentially as described by Lagunas et al.4, built from the PDB structure 3CX573, using the Multiseq plugin74 of the VMD program75. The two proteins were separated 3.2 nm along one axis, taking the most external hydrogen atom from the -CH3 distal pairs from each heme group, that is 4.2 nm taking the Fe-to-Fe distance. The molecular dynamics (MD) simulation was set up with the program LEaP included in the Amber suite of programs76 and the ff99SB protein force field. The titratable residues were modeled in the corresponding protonation state at pH 6.5, which was further checked by analysis of their intermolecular interactions. The AMBER parameters database (http://research.bmh.manchester.ac.uk/bryce/amber) was used for the heme parameters, considering RESP charges for the reduced and oxidized cases from Autenrieth et al77. Values from the same reference were used to parametrize the charges of the interacting residues with the heme groups, His18/44 and Met80/166 from Cc/Cc1 respectively. Explicit TIP3P water molecules were used for solvation78, and the system was neutralized with 33/35 sodium/chlorine atoms, in addition to the 33 NaCl molecules to reproduce a concentration of 50 mM. The final system contained 105541 atoms. Amber1479 was used for the MD simulations. A restraint on the residues 1–15 for Cc1 and 97–104 for Cc was applied to model the experimental situation of Cys10 and Cys104, respectively, immobilized on the gold surfaces. A thermal equilibration to 300 K was first done, followed by the equilibration dynamics in the NPT ensemble for 36 ns. We used the SHAKE algorithm, with an integration time step of 2 fs. To obtain the equipotential surfaces, APBS electrostatics calculations80 were done in the postprocessing step. Average ions and water concentration maps and number density prism volumes were obtained with the Volmap Plugin 1.1. Matplotlib81 and VMD75 were used to draw the figures.

Brownian dynamics simulations

Brownian dynamics (BD) computations were carried out and analyzed using the SDA-flex 7.1 software package82. The force-field grids used in BD included all electrostatics and desolvation65,83 grids. Charges were obtained from PQR files extracted from each MD trajectory. Diffusion constants were computed using the script ARO84, in the VMD tcl-tk console. Electrostatic grids were generated for every conformer with APBS 3.085. All simulations were carried out at 100 mM ionic strength. In total, 20 structures of Cc1 and other 20 of Cc were used as input. For kon computations, each of the 20 Cc1 molecules were treated as targets in separate computations and set in the coordinate origin, whereas the 20 conformations of Cc were used as input for Monte-Carlo conformation exchange during the simulation of their diffusion. Conformational exchange was allowed every 2.5 ns. A total of 200,000 diffusion trajectories (20 × 10,000) were then computed. Origin 2019b (Originlab) was used for statistical analysis and data representation.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.