Electron transfer in an acidophilic bacterium: interaction between a diheme cytochrome and a cupredoxin

Electrochemical studies of diheme cytochrome/cupredoxin complexes provide new insights into the electron transfer pathway in an acidophilic bacterium.


Introduction
Biodiversity is an extraordinary source of microorganisms displaying unusual features such as resistance to high temperatures, high pressures, high salinity, extreme pHs, etc. Although these organisms were identied many years ago, the molecular factors that allow them to survive and grow under such extreme conditions are far from being completely understood. Extreme acidophiles grow optimally at an external pH of 3 or less. Their ability to survive such drastic environments has aroused a great deal of interest, from both a fundamental and an application point of view. 1 Acidithiobacillus ferrooxidans is one of the best studied model organisms and has been used for a long time in bioleaching mine processes and more recently in microbial fuel cells, thanks to its ability to gain energy through the oxidation of ferrous iron at pH as low as 2, thus allowing overcoming the issue of proton availability required to construct an efficient biocathode. 2,3 It has been clearly shown that Fe 2+ oxidation takes place outside the bacterial cell, while the electrons generated are driven into the periplasmic compartment via electron shuttling to the inner membrane where O 2 reduction takes place. 4 Several metalloproteins involved in this metabolic chain have already been isolated (Scheme 1), and details on the components have been reported in comprehensive reviews. 5,6 The primary electron acceptor, cytochrome Cyc2, is an outermembrane monoheme c-type cytochrome. Then periplasmic proteins, including a blue copper protein, rusticyanin (Rus), and a diheme c-type cytochrome (Cyt c 4 ), have been shown to transfer electrons to the terminal electron acceptor, cytochrome c oxidase (CcO), an inner-membrane protein belonging to the subgroup of heme-copper O 2 reductases. 7 All these proteins have in common a high value of their redox potentials, 5 which ts with the redox potential of the Fe 3+ /Fe 2+ redox couple at acidic pHs. 8 A docking model suggests that Cyt c 4 acts as a wire between Rus and CcO, with the heme of the highest potential (Heme H ) in interaction with Rus, and the heme of the lowest potential (Heme L ) in contact with CcO. 9 Nevertheless, different models have been proposed to describe the full electron transfer (ET) pathway, and divergence exists concerning the sequential ET pathway between the proteins in the chain. [10][11][12] Another copper binding protein, AcoP short for "Acidophile Cytochrome c Oxidase Partner", has been found to be directly involved in the respiratory chain. 13 It has been shown that AcoP copuries with CcO, an enzyme which, in other organisms, is well known to receive its electron from a soluble cytochrome c. Most studies have consequently assumed a direct ET between Cyt c 4 and CcO in A. ferrooxidans, not taking into account AcoP. 6,12 The presence of AcoP as an additional copper metalloprotein in tight interaction with CcO raises the question of its role in the respiratory chain. Its genetic organization conrms its involvement in the respiratory chain as the acoP gene is in an operon with the genes encoding the entire pathway. 5 The presence of a copper site in AcoP may suggest a function beyond the stabilization effect on CcO activity in an acidic environment. 13 Preliminary observations proposed a potential interaction between puried AcoP and Cyt c 4 . 13 In addition, AcoP's intrinsic properties have recently been fully characterized and point out that it is a novel member of the cupredoxin family with a green copper centre, also called a type 1.5 copper centre, displaying the highest redox potential described to date for this subclass (+350 mV at pH 4.8 vs. Ag/AgCl). 14,15 On the basis of its high redox potential and its cupredoxin fold, well suited for optimized ET, a role of AcoP as an electron shuttle might be envisioned, although it has not been demonstrated so far.
Electrochemical techniques are the methods of choice to study the thermodynamics and kinetics of ET processes involved in energy chains. 16,17 ET between various types of electrodes, including carbon or gold surfaces, and small redox proteins such as heme, 18-20 copper 21 or iron-sulfur clusters 22 containing proteins, was investigated early on to determine the key factors allowing the electron exchange at the interface. Intramolecular ET between several redox sites in the same protein has also been reported. In the cases where two redox sites are present, thanks to a specic orientation of one site over the other on the electrode, specic electrochemical signals have helped in the quantication of both the interfacial and the intramolecular ET. 23,24 These preliminary studies have served the further quantication of the intermolecular ET within the transitory complex formed between an enzyme and its physiological partner. As examples among many others, second order rate constants between polyheme cytochrome c 3 and hydrogenases, 25,26 or nitrite reductases 27 and azurin 28 or cytochromes 29 have been obtained by recording the current response to a potential step at electrochemical interfaces. Beyond this, all these studies opened the way for the study of direct catalysis using enzymes at electrochemical interfaces. 30,31 In this paper, we focus on the interaction between AcoP and one of its putative partners in the ET chain, Cyt c 4 , two proteins operating in the pH range of 2.5-3 in the periplasmic compartment. 5 To our knowledge, very few electrochemical studies have been reported on the interaction between two proteins interacting in an ET pathway even though not directly involved in an enzymatic process. One relevant but rare example is the early voltammetric study of the intermolecular ET between cytochrome c and cytochrome b 5 , two proteins shown to interact through a protein-protein ET complex. 32 Thanks to the purication to the homogeneity of both AcoP and Cyt c 4 from A. ferrooxidans, we now have the opportunity to investigate an additional ET chain by electrochemistry. In the work presented here, we rst used biochemical techniques to demonstrate the occurrence of a complex between AcoP and Cyt c 4 which allows ET in a homogeneous aqueous phase. To reveal the molecular determinants of a protein-protein interaction favourable to ET in the acidophilic chain of A. ferrooxidans, we then analysed the interfacial ET between each individual protein and an electrode, with the aim of mimicking partnerpartner interaction. For this purpose, the surface chemistry of the electrode was modied and further modulated by pH and ionic strength. We nally took advantage of the knowledge of the electrode surface chemistry required for the interfacial ET on an individual protein, with the ultimate goal of analysing the consequence of the formation of a complex between AcoP and Cyt c 4 on the electrochemical signal. We discuss the occurrence Scheme 1 Iron respiratory chain of Acidithiobacillus ferrooxidans. of intermolecular ET between the two proteins at the electrochemical interface. This ET process suggests new insights into the ET pathway allowing A. ferrooxidans to grow under acidic conditions.

Results and discussion
Complex formation and intermolecular electron transfer between Cyt c 4

and AcoP in solution
To demonstrate the formation of a complex between AcoP and Cyt c 4 , we rst puried both proteins to homogeneity as shown in Fig. 1.
Both puried proteins were incubated for 12 hours, either separately or together, and subsequently loaded on a "modiednative" gel containing a small amount of SDS (0.0175% versus 0.1% in SDS-PAGE gel). As shown in Fig. 1C, in this experimental set-up, a clear additional band, of a higher molecular weight than the individual proteins, and labeled (*) could be visualized when AcoP and Cyt c 4 were incubated together. This additional band was not seen using SDS-PAGE gel (ESI Fig. S1 †). To conrm the presence of both proteins in the band labeled (*), N-terminal sequencing was performed. The results presented in ESI Fig. S2 † unequivocally conrm the presence of AcoP and Cyt c 4 and thus the existence of an observable AcoP/ Cyt c 4 complex. No complex formation occurred between AcoP and Cyt c 4 in the presence of ascorbate, suggesting that it is dependent on the redox state of the proteins (Fig. 1C).
The involvement of AcoP in an electron pathway through Cyt c 4 in solution was then evaluated. Puried Cyt c 4 in its reduced form ( Fig. 2A) was incubated with "as prep" AcoP in an equimolar amount. UV-vis spectroscopy clearly showed Cyt c 4 oxidation in the presence of AcoP, revealed by the decrease of the b and g bands of heme c groups at 523 and 552 nm, and a net change of the ratio between these two bands ( Fig. 2A). For comparison, the spectrum of Cyt c 4 reduced by ascorbate and oxidized by K 2 IrCl 6 is shown in the Fig. 2A inset.
Given the low absorption coefficient of AcoP at 570 nm ($2500 M À1 cm À1 (ref. 15)) compared to that of Cyt c 4 at a similar wavelength (3 552 nm ¼ 46 000 M À1 cm À1 (ref. 33)), the concomitant reduction of AcoP cannot be observed using this approach. To observe AcoP and Cyt c 4 signatures independently, EPR measurements of frozen solutions of Cyt c 4 , AcoP and AcoP/ Cyt c 4 1 : 1 mix were performed (Fig. 2B). The Cyt c 4 spectrum exhibits two positive peaks, at g ¼ 3.39 and 3.13, corresponding to the g z -values of oxidized Heme H and Heme L , respectively. 34 The broad signal around g ¼ 2.3 can be attributed to the g y parts of heme signals. As expected for low-spin ferric heme signals with a maximum g value higher than 3, the g x contributions are  too broad to be detected. 35 The AcoP spectrum was characteristic of a rhombic Cu(II) centre with g-values equal to 2.193, 2.057 and 2.019, and copper hyperne coupling constants of 66 Â 10 À4 , 12 Â 10 À4 and 65 Â 10 À4 cm À1 respectively. 14 The addition of one equivalent of Cyt c 4 led to a signicant decrease of the AcoP signal, suggesting the reduction of the Cu(II) site. In addition, the Heme H peak at g ¼ 3.39 increases by a factor of 2.4 in intensity, which indicates an oxidation of the Heme H concomitant with the reduction of the AcoP copper site. Thus UV-vis and EPR spectroscopies both demonstrate that ET between AcoP and Cyt c 4 occurs in solution in vitro, most probably through Heme H .

Interfacial electron transfer properties of AcoP
More insights into the conditions and kinetics of complex formation between AcoP and Cyt c 4 were obtained by using electrochemical techniques. The additional interest of using electrochemistry was that the electrochemical interface could be easily tuned to give access to the key parameters allowing the relationship between interfacial ET and pH, ionic strength, etc., to be investigated. The data obtained by such an investigation could help to dene the type of interaction involved in the interfacial ET between the electrode and each protein, and, beyond, between the two proteins if we consider the electrode as one of the partners. We rst mimicked the AcoP-partner interaction by studying the electrochemical behavior of AcoP under different experimental conditions at two different electrodes, pyrolytic graphite (PG) and gold, the electrode mimicking its partner, Cyt c 4 .
The role of potential electrostatic interactions was rst evaluated by studying the electrochemical behavior of AcoP at a PG electrode. Three particular pHs, 3.5, 4.8 and 7, were chosen as they allowed both the global charge of the protein and the charge of the electrode to be tuned. The PG surface carries many carboxyl functions that induce a pK a close to 5. 36 Modifying the electrode at these different pHs enabled positively (pH 3.5) or negatively charged (pH 7) electrodes to be created. AcoP has a theoretical pI of 7.2 and is thus globally positively charged at pH 3.5 and 4.8, and neutral at pH 7. Fig. 3 reports the voltammetric behavior of AcoP entrapped in the thin layer of the membrane electrode conguration at these 3 pHs in 20 mM NH 4 AC buffer before and aer the addition of 200 mM NaCl, which was used to screen possible electrostatic interactions. 37 The relationship between the mean potential and pH is given in ESI Fig. S3. † At pH 4.8, the one-electron redox wave characteristic of the Cu site in AcoP appears with an average potential of +350 AE 5 mV, in agreement with previous measurements either in solution using UV-vis titration 14 or at a PG electrode. 15 The difference between the cathodic and the anodic peaks, DE p , is close to 120 mV at 20 mV s À1 , denoting a slow ET rate. The addition of up to 200 mM NaCl in the buffer did not induce a change either in the peak current or in DE p . When cyclic voltammetry (CV) was performed at pH 7, a redox wave at +340 AE 5 mV was observed displaying a DE p of 90 mV, which was not stable over consecutive cycles (ESI Fig. S3 †). The addition of NaCl tended to increase DE p to 160 mV, but also induced the stability of the signal. Decreasing the pH to 3.5 or lower led to a drastic loss of the reversible redox wave, which hardly reappeared aer NaCl addition. The transfer of the AcoP-based electrode from pH 3.5 back to pH 4.8 did not lead to the recovery of the redox signal.
The evolution of the CV signal of AcoP as a function of buffer pH and ionic strength can be explained considering the distribution of protein charges at a given pH (Fig. 4), which induces dipole moments of 269, 216 and 145 Debye at pH 3.5, 4.8 and 7, respectively. Since we used a model of the AcoP structure, the exact dipole moments might be slightly different, but we believe that the trend is realistic. These dipole moments are 3-4 times lower than values obtained for proteins known to interact with the electrode surface through electrostatic interactions, 38,39 suggesting that this type of interaction is not the main force driving the interaction of AcoP with the electrode. Accordingly, interfacial ET occurs only under conditions where electrostatic interactions are weak, such as pH 4.8, where AcoP is globally positively charged while PG is neutral, or pH 7, where the protein is globally neutral and the electrode is negatively charged. In the latter case, ET is stabilized by the presence of NaCl, which weakens electrostatic interactions. The loss of the redox wave at low pHs was unexpected for an acidophilic protein and raises the question of AcoP structural stability on the PG electrode. UV-visible spectra had previously shown that copper geometry was not affected by changes in pH in the range from 3.5 to 7.4. 14 The present work further demonstrates that even a pH as low as pH 2.8 does not induce any modication of the copper centre (ESI Fig. S4 †). Circular Dichroïsm (CD) (Fig. 3D) and ATR FTIR (Fig. 3E) measurements of AcoP solutions at different pHs conrm that the decrease of pH has no effect on the secondary structure of AcoP. CD spectra can be superimposed in the pH range of 2.8-7, and no shi of the amide I band at around 1630 cm À1 , characteristic of a structure rich in b-sheets, was observed in the ATR FTIR spectra.
UV-vis, CD and ATR-FTIR measurements underlined that AcoP was able to tolerate a broad range of pH conditions in solution, with no impact on its overall fold. These observations suggest that the absence of interfacial ET at low pH is most probably linked to the interaction with the electrode surface, which irreversibly impacts the native conformation of AcoP. Polarization modulation-infrared reection-adsorption spectroscopy (PMIRRAS) on a Self-Assembled-Monolayer (SAM) on gold surfaces was used to study any conformational change of the protein upon immobilization on the electrode, especially at low pH. A 6-Mercaptohexanoic acid (MHA)-based SAM (Fig. 5A) was chosen because the induced surface chemistry and pK a (pK a of MHA as a SAM is close to 6 (ref. 18)) resemble those of PG. The carboxylic groups of the MHA-based SAM are observed at around 1735 cm À1 . Both amide I and amide II bands appear on PMIRRAS spectra whatever the pH (Fig. 5A), demonstrating that AcoP is readily adsorbed on MHA-based SAMs at the two pHs under investigation. However, the position of the amide I band differs as a function of pH, shiing from 1655 cm À1 to 1670 cm À1 at pH 4.8 and 2.8, respectively. The typical band at 1655 cm À1 corresponds to the amide group involved in a-helices, while the band at 1670 cm À1 is assigned to the amide group implicated in turns. As described above, AcoP in solution is mainly structured in b-sheets, with a characteristic wavenumber of 1630 cm À1 . The rst conclusion is that adsorption on MHA-based SAMs induces either a modication of the protein secondary structure or a specic orientation of AcoP with the main contribution of the AcoP a-helix perpendicular to the  surface according to PMIRRAS selection rules. 40 At pH 4.8, because AcoP is still electroactive (Fig. 5B), strong denaturation is unlikely, but instead the maximum observed at 1655 cm À1 would reveal an orientation of the AcoP on the surface. At pH 2.8, the AcoP amide I band was shied from 1655 cm À1 to 1670 cm À1 . At this pH, the loss of electroactivity (Fig. 3) would be favorable for some modication of the secondary structure rather than variation in the orientation of AcoP. In conclusion, AcoP adsorption on hydrophilic surfaces at low pH affects its secondary structure, and hence interfacial ET.
Native AcoP has a transmembrane segment and has been proposed to interact with the membrane bound CcO. Having ruled out a major role of electrostatic interactions, we proceeded to test whether hydrophobic interactions could rather drive interfacial ET and consequent complex formation. Fig. 4 underlines some hydrophobic patches on the surface of the protein which could be involved in the partner recognition. In a mimicking way, the putative recognition of AcoP with the electrode surface through hydrophobic interactions was studied on a hydrophobic butane thiol (BT)-based SAM on a gold electrode, on which fast ET on azurin was previously obtained. 28 No SWV signal was observed with AcoP adsorbed on the BT-SAM (Fig. 5B). PMIRRAS was used to compare the conformation of the protein upon immobilization on the BT-SAM with that on MHA-based SAM surfaces (Fig. 5C). The spectra show that AcoP is also readily adsorbed on a BT-based SAM surface. Changes in the conformation of AcoP can be observed, with a loss of b sheets. The amide I position is shied to 1670 cm À1 , revealing the presence of a high level of turns in the structure of AcoP adsorbed on the hydrophobic surface. AcoP adsorption on a BTbased SAM electrode might induce a conformational change of the protein, explaining the absence of an electrochemical response on a hydrophobic surface.
The conclusions that can be drawn from electrochemistry and PMIRRAS are that hydrophobic surfaces, or strong electrostatic force between highly positively charged proteins at low pH and hydrophilic surfaces, irreversibly modify the AcoP structure and impair ET to the electrodes. This is in agreement with the fact that periplasmic proteins from acidophiles operate at pH lower than 3 and are highly positively charged (Fig. 4). Under such conditions, electrostatic interactions mostly involve the repulsion between positive charges impairing the ET. This would explain the requirement of weak electrostatic interactions between AcoP and the electrode for efficient ET. It cannot be excluded that hydrophobic interactions might drive the interaction between proteins in such an environment, but would involve small hydrophobic patches. A novel design of electrodes would be required to mimic such surfaces.

Interfacial ET properties of Cyt c 4
To further evaluate the factors that can control protein-protein interaction in the metabolic chain of A. ferrooxidans, the electrochemical behavior of Cyt c 4 was studied at the three pHs used previously, i.e. pH 3.5, 4.8 and 7, with a PG electrode acting as a partner (Fig. 6). At all these pHs, Cyt c 4 is globally positively charged (the theoretical pI of Cyt c 4 is 8) (Fig. 7). At pH 4.8 ( Fig. 6A), two redox waves were observed, with mean potentials of +235 AE 5 mV and +125 AE 5 mV, corresponding to the high potential heme, Heme H , and the low potential heme, Heme L , respectively. These values are in good agreement with the previous ones determined at a gold electrode modied with bis(4-pyridyl)disulde. 9 The ratio of peak heights of Heme H and Heme L is close to 1, and DE p for both processes is around 30 mV at 20 mV s À1 , showing that the ET rate is fast and equivalent for the two hemes. As with AcoP, the addition of NaCl did not change the prole of the CV. When the membrane was removed from the electrode, the redox waves for Cyt c 4 persisted (Fig. 6D), showing the stable adsorption of some molecules. From the charge under the redox waves at different scan rates, a surface coverage of 20 AE 2 pmol cm À2 was calculated. At more basic and more acid pHs ( Fig. 6B and C), the curves for the two redox processes for Heme H and Heme L were still well shaped. Notably at pH 7, the redox signal for Cyt c 4 was not stable over time, even aer the addition of NaCl (ESI Fig. S5 †). Using CD and optical spectroscopy we did not see any critical modication of the Cyt c 4 structure in solution in the pH range of 2.8-7 (Fig. 6E, ESI  Fig. S6 †). Cyt c 4 loss of activity on the electrodes at pH 7 might have been due to electrostatic interaction with the negatively charged electrode surface which was far from physiological conditions for acidophilic organisms.
Overall, the occurrence of a redox signature for both hemes, and their equal intensities whatever the pH, and hence whatever the charge of the PG electrode, suggest that Cyt c 4 cannot be specically oriented at the electrochemical interface. This is different from what was found for the Cyt c 4 from Pseudomonas stutzeri, which had distinct positive and negative domains surrounding each heme at pH 7 (ESI Fig. S7 †), properties that allowed its orientation on a negatively charged SAM electrode with one heme facing the electrode. 23 The absence of such an orientation of Cyt c 4 was further assessed by immobilizing the protein on SAM electrodes bearing different charges (Fig. 6F). Well-shaped SWV signals were obtained at negatively charged MHA and positively charged 4-aminothiol (4-ATP) SAM electrodes at pH 4.8, with redox potentials of 230 AE 5 mV and 115 AE 5 mV, corresponding to Heme H and Heme L . Cyt c 4 from A. ferrooxidans displays a dipole moment ranging from 256 D to 420 D depending on the pH, half that of P. stutzeri Cyt c 4 at pH 7. A comparison of their surface charges (ESI Fig. S7 †) claries the non-orientation of A. ferrooxidans Cyt c 4 , as positive and negative charges are randomly exposed to the protein surface, in contrast with P. stutzeri Cyt c 4 . Interestingly, on a hydrophobic BT-based SAM, the redox waves relative to the two hemes also appeared at their expected redox potentials, demonstrating an efficient ET interaction between Cyt c 4 and hydrophobic surfaces. This interaction can be easily understood by considering the large hydrophobic areas on the surface of the protein (Fig. 7). This is also in agreement with molecular dynamic simulations that showed that the interaction between Cyt c 4 and CoxB, the subunit of CcO transferring the electrons to the catalytic site, was stabilized by hydrophobic residues. 12 Unlike AcoP, Cyt c 4 thus appears as a versatile partner that should afford many types of interactions for ET with both hemes. In line with this conclusion, it was recently proposed that Cyt c 4 from A. ferrooxidans would be involved in the anaerobic electron transfer chain coupling Fe 3+ reduction to sulfur oxidation. 41 In such a case, Cyt c 4 would transfer electrons to other partners than CcO. Interestingly, versatility has also been demonstrated recently using the diheme cytochrome c 550 from Thermus thermophilus. 42 Whether it is a general property of diheme cytochromes is an interesting open question.

Electrochemical behavior of a mixture of AcoP and Cyt c 4
Thanks to the knowledge of the electrochemical behavior of individual proteins, the kinetics of ET between AcoP and Cyt c 4 was further studied using electrochemistry. Fig. 8 shows the CVs obtained at 20 mV s À1 at the PG electrode in the membrane conguration with a mixture of AcoP and Cyt c 4 in a 1 : 1 ratio, incubated for 12 h at pH 4.8. The rst CV cycle was run in the forward scan from À0.1 V to +0.6 V. The second cycle superimposed onto the rst one (not shown). The signals for AcoP and Cyt c 4 alone under the same experimental conditions are overlaid for comparison. Interesting features can be extracted from these CV experiments. Redox waves corresponding to Heme L in the mixture are similar to those found with Cyt c 4 alone, both in terms of mean potential and peak height. Such a result demonstrates that the low potential Heme L is not affected by the presence of AcoP. The oxidative wave for Heme H is also conserved in the mixture compared to Cyt c 4 alone. In contrast, no well-shaped redox signal at the expected potentials for AcoP can be observed in the CV of the mixture, suggesting that heterogeneous interfacial ET for AcoP is slowed down in the presence of Cyt c 4 . However, a reduction wave at a peak potential around +250 mV appears with a marked increase in the peak current compared to either Heme H of Cyt c 4 or AcoP alone.
The reduction wave at +250 mV, denoted peak ① in the following, strongly decreases when the forward scan of the CV is restricted to +350 mV, i.e., before the expected oxidation of AcoP (Fig. 8B). Peak ① thus involves the reduction of oxidized AcoP. Peak ① does not appear when the same experiment is performed with Horse heart Cyt c instead of Cyt c 4 (Fig. 8C). This last result strongly suggests that peak ① is linked to a specic interaction between AcoP and Cyt c 4 . We repeated similar experiments with the mutant AcoP His166A previously shown to have lost its redox properties, even though its secondary structure and copper content were unmodied (Fig. 8D). 13 Peak ① was absent when AcoP His166A was incubated with Cyt c 4 , underlining the importance of a redox active state of AcoP to observe the particular electrochemical behavior of the complex.
In a study focusing on the interaction between Cyt c and cytochrome b 5 (Cyt b 5 ), CVs of a mixture of these two redox proteins revealed a pre-peak before the reduction signal for Cyt b 5 . It was proposed that this reductive process was due to fast intermolecular ET between reduced Cyt b 5 and oxidized Cyt c. Actually, the electrochemical interface was chemically modied so that Cyt c could be reduced at the interface only through the electrogenerated reduced form of Cyt b 5 . 32 The situation is somewhat different in the case of AcoP and Cyt c 4 because the interfacial ET for AcoP is hindered but not completely forbidden. Effectively, oxidation current is noticeable in Fig. 8A in the potential range for AcoP oxidation, which can be attributed to slow interfacial ET. Nevertheless, the typical CV signal we have observed with the protein mixture can also be attributed to intermolecular ET through the complex formed between Cyt c 4 and AcoP, whose formation is not driven by electrostatic interactions. Indeed, the addition of 400 mM NaCl to the electrolyte did not drastically modify the CV (Fig. 8E). Peak ① is thus the result of intermolecular ET between electrogenerated reduced Cyt c 4 and oxidized AcoP according to the processes illustrated in Scheme 2. Since we demonstrated that Heme L does not participate in the interaction with AcoP, its independent electrochemical processes are excluded from the ET model. To explain the full CV, we propose the following kinetic steps: (i) complex formation between the oxidized forms of the protein, (ii) heterogeneous reduction of Cyt c 4 Heme H through favoured interaction of the complex with the PG electrode via the versatile adsorbed Cyt c 4 , (iii) reduction of AcoP through intermolecular ET from the reduced Heme H of Cyt c 4 along peak Scheme 2 Electrochemical and chemical steps involved in the ET pathway from the oxidized form of the AcoP and Heme H Cyt c 4 complex to the reduced form explaining the appearance of peak ① observed in Fig. 8. ①, (iv) dissociation of the complex aer the reduction of the proteins, (v) heterogeneous oxidation of Cyt c 4 , (vi) slow oxidation of AcoP at potentials above +350 mV, and (vii) reassociation of the two oxidized proteins through the complex.
If the peculiar electrochemical behavior of the mixture of Cyt c 4 and AcoP is linked to the formation of a complex that facilitates AcoP reduction, the sweep rate should affect the overall signal. To gain more insight into the redox processes involving the AcoP : Cyt c 4 mixture, the CVs at 20 mV s À1 were compared to CVs obtained at 2 and 200 mV s À1 . Typical CVs are shown in Fig. 9, where for each sweep rate, the signatures for AcoP and Cyt c 4 alone and for the mixture are overlaid. At 2 mV s À1 (Fig. 9A), the CV of the complex is characterized by two redox waves assigned to Cyt c 4 , with a decrease in the peak height proportional to the sweep rate, as expected for thin layer electrochemistry. As the interfacial ET for AcoP alone is a slow process, the oxidative signal for AcoP at a low scan rate is now visible at the expected potential value. Nevertheless, peak ① is still observed, meaning that intermolecular ET is faster than the heterogeneous reduction of AcoP. The experiments described before were carried out at pH 4.8. In the context of an acidophilic chain, it should be more relevant to study the complex electrochemical behavior at a physiological pH. Thus, we recorded the CVs at pH 2.8, in the presence of the AcoP : Cyt c 4 mixture at the PG membrane electrode. Peak ① linked to the reduction of AcoP via intermolecular ET through Cyt c 4 was clearly observed (Fig. 9B). As the interfacial ET for AcoP alone is very slow at this low pH, AcoP reduction can only proceed through intermolecular ET via Cyt c 4 . This is an additional proof of the occurrence of an efficient ET complex between the two partners.
At 200 mV s À1 , only the redox waves for Cyt c 4 are observed (Fig. 9C). At this high speed, one or several kinetic processes may be limiting steps precluding the peak ① observation: slow AcoP oxidation, slow complex formation or slow intermolecular electron transfer within the complex. To verify the inuence of each process and conrm complex formation, numerical modeling of the CVs as described in the ESI (Fig. S8 †) was performed at 2, 20 and 200 mV s À1 as a function of the heterogeneous electron transfer constant of AcoP (k 0 ), the kinetic rate constant of the complex formation (k as ) and the constant of the intermolecular electron transfer within the complex (k inter ). Fig. 10 illustrates one such simulation at 20 mV s À1 .
The shapes of the theoretical CV curves tend to match those of the experimental ones, conrming the intermolecular ET through a complex between adsorbed Cyt c 4 and AcoP. The constant of the intermolecular electron transfer has the strongest inuence on the potential of peak ① which only slightly varies as a function of the other two constants. This can be used for the estimation of this constant. Based on the simulation at the three different scan rates, and hypothesizing a fast complex formation which is thermodynamically favoured, the intermolecular electron transfer within the complex could be estimated to occur with a rate of 3-5 s À1 . This value is smaller than other values estimated from photoinduced intracomplex ET between Cyt f and Cyt c 6 from Chlamydomonas reinhardtii (9 Â 10 2 s À1 ) 43  or intracomplex ET obtained by laser excitation of rutheniumlabeled Cyt b 5 and cyt c (4 Â 10 5 s À1 ). 44 Our intermolecular ET between AcoP and Cyt c 4 is in the same range as the value reported for ET between Pseudomonas aeruginosa cyt c 551 and Silene pratensis plastocyanin obtained by stopped-ow measurements (5.7 s À1 ). 45 Two hypotheses may explain the slow intermolecular ET rate between AcoP and Cyt c 4 : (i) the adsorption of Cyt c 4 at the electrode surface impairs the optimal ET complex; (ii) AcoP has additional functions other than accepting electrons from Cyt c 4 .
Recombinant AcoP without its transmembrane segment was expressed and puried from Escherichia coli as described previously. 14 AcoP protein concentration was determined by UVvis absorbance spectroscopy, measuring the absorption intensity at 280 nm and using a sequence-derived theoretical molar extinction coefficient of 25 440 M À1 cm À1 . Puried AcoP was stored in 50 mM NaAC buffer at pH 4.8, containing 0.005% (w/v) DDM. When mentioned, AcoP was dialysed against 10 or 20 mM NH 4 AC buffers adjusted to different pH values, pH 2.8, pH 3.5, pH 4.8 and pH 7.
Cyt c 4 was expressed as described previously in 100 l E. coli cultures 9,34 using three separation steps on carboxymethyl cellulose, SP-Sepharose and monoS columns. The enriched fractions containing Cyt c 4 in 80 mM NH 4 AC buffer at pH 4.8 were pooled and dialysed against 80 mM NH 4 AC buffer at pH 2.8, followed by another dialysis at pH 4.8. This pH variation allowed further elimination of E. coli acid unstable proteins. When mentioned, puried Cyt c 4 was dialysed against 10 or 20 mM NH 4 AC buffers adjusted to different pH values, pH 2.8, pH 3.5, pH 4.8 and pH 7.
Cytochrome c (Cyt c) from an equine heart ($95%, SDS-PAGE) was bought from Sigma-Aldrich. Cyt c was diluted in NH 4 AC 20 mM buffer at pH 4.8.

Protein modelling
For cupredoxins, being well known, several structures have been solved which serve as templates for modelling the globular soluble domain of AcoP. Here we used the Phyre2, Swiss-model and I-TASSER servers 47 obtaining similar results. We note that, given the high homology with the templates, this model is expected to be of reasonable quality along most of the sequence. 48,49 For Scheme 1, AcoP's TM helix was built manually by threading its sequence as an ideal helix. The model of Cyc2 in Scheme 1 is based on I-TASSER homology modelling guided by coevolution-based restraints as described in ref. 50. For Cyt c 4 of Acidithiobacillus acidophilus (pdb: 1H1O) and Cyt c 4 of Pseudomonas stutzeri, we used available X-ray structures (pdb: 1H1O and 1ETP, respectively).
To study protein electrostatics, AcoP models and Cyt c 4 structures were prepared with the PDB2PQR-Propka servers, using the Amber force eld to assign electrostatic charges at the surface of the proteins and the Parse force elds to determine the dipole moments of proteins with PDB2PQR-Propka servers. 51 Dipole moments were also determined from the pqr output les using the Protein Dipole Moments Server. 52 Electrostatic charges at the surface of proteins and surface hydrophobicity were illustrated using PyMOL (The PyMOL Molecular Graphics System, Version 1.8 Schrodinger, LLC).
Modied-native-PAGE analysis. Samples were mixed with DDM at a nal concentration of 0.01% and incubated at room temperature for 45 min. The treated samples were mixed with 2Â modied Laemmli buffer (1Â contains 125 mM Tris (pH 6.8), 25% glycerol, 0.0175% SDS, and 0.01% bromophenol blue) and charged directly on 15% Tris-glycine native gel containing 0.0175% SDS. The gels were run in Tris-glycine running buffer with 0.0175% SDS. Proteins were visualized by Coomassie Blue or PageBlue staining.
N-Terminal sequence determination. N-Terminal sequence determination was performed by stepwise Edman degradation using an automatic sequencer model PPSQ 31B (Shimadzu, Kyoto, Japan).

Electrochemical measurements
Cyclic voltammetry (CV) and square wave voltammetry (SWV) were performed using an Autolab PGSTAT30 potentiostat analyzer controlled by Nova soware (Eco Chemie). The electrochemical cell was equipped with three electrodes, a pyrolytic graphite (PG) or a gold electrode as the working electrode (S ¼ 0.07 cm 2 ), a platinum wire as an auxiliary electrode and an Ag/ AgCl electrode as the reference electrode. All potentials are quoted vs. the Ag/AgCl reference electrode here. Potentials versus NHE can be obtained by adding 210 mV to the reported potentials. 20 mM NH 4 AC buffer was used as the electrolyte. All the electrochemical measurements were made at least in triplicate at 25 C.
The PG electrode surface was renewed by polishing with ne sand paper (P1200), and then briey sonicated to remove free carbon particles. The membrane electrode conguration was used to entrap 2 ml of protein sample in a thin layer between the electrode and a dialysis membrane of suitable cutoff. 53 The gold surfaces were cleaned with "piranha" solution (3H 2 SO 4 98% : 1H 2 O 2 30%) for 4 min and rinsed extensively with water and later with ethanol. Self-assembled monolayers (SAMs) were formed by immersing the gold surfaces in 5 mM ethanol solutions of 4-ATP or MHA or BT for 18 hours. The surfaces were then cleaned with ethanol to remove all organic contaminants.

Spectroscopies
Circular Dichroism (CD) spectra of 15 mM proteins were recorded on a Jasco J-715 spectropolarimeter at 298 K in a 1 mm path length cell. Spectra were averaged from ve scans and normalized for any variation in protein concentrations measured at 280 nm by optical spectroscopy using a Cary 50 Bio (Varian).
UV-vis spectra were recorded on a Cary 50 Bio (Varian) spectrophotometer. Protein samples (AcoP or Cyt c 4 ) were used at 15 mM, for a scan recorded from 200 nm to 1000 nm to follow metal center properties. When mentioned, oxidation or reduction of the samples was performed using an equimolar amount of protein and potassium hexachloroiridate(IV) (K 2 IrCl 6 ) or ascorbate. Ascorbate can be removed for further analysis by using a desalting column (PD-10 column, GE Healthcare). Because iridium is a strong oxidizing agent that irreversibly modies key residues, we used "as prep" AcoP which corresponds to $40% of the oxidized form. 14 X-band electron paramagnetic resonance (EPR) spectra were recorded using a Bruker-Biospin EleXsys E500 spectrometer equipped with a standard rectangular Bruker EPR cavity (ER4102ST) connected to an Oxford Instruments helium ow cryostat (ESR900).
For polarization modulation-infrared reection-adsorption spectroscopy (PM-IRRAS), the modied dried gold surface was placed at room temperature in the external beam of the FT-IR instrument on a Nicolet Nexus 870 FT-IR spectrometer (Madison, WI). The experimental set-up has been reported in a previous paper. 54 The PMIRRAS spectra were recorded at 8 cm À1 resolution, with the co-addition of six hundred scans.
ATR-FTIR spectra were recorded on a Nicolet 6700 FT-IR spectrometer (Nicolet Instrument, Madison, WI) equipped with a liquid nitrogen cooled mercury-cadmium-telluride detector (Thermo Fisher Scientic, San Jose, CA), with a spectral resolution of 4 cm À1 . Two hundred interferograms were coadded.

Conclusions
In this study, we demonstrated the formation of a complex between AcoP and Cyt c 4 , two proteins of an acidophilic bacterial respiratory chain. We used an electrode interface to mimic AcoP and Cyt c 4 as putative redox partners in the energy chain. We found that electrostatic interactions do not drive efficient ET from the AcoP copper centre to the electrodes. In contrast, we showed that Cyt c 4 is highly versatile, displaying ET at the two heme redox centres on varied types of surfaces, either positive, negative or hydrophobic. This suggests that Cyt c 4 could be a hub that controls the rate of electrons towards different routes, depending on the metabolic or environmental states. We also demonstrated an intermolecular ET between AcoP and Cyt c 4 , allowing the reduction of AcoP through the high potential heme of Cyt c 4 (Scheme 3). Thanks to modeling of the electrochemical signals, we were able to quantify the intermolecular ET rate within the complex. Although the results for the mixture suggest a likely directionality of ET from Cyt c 4 to AcoP, the value of the intermolecular rate constant opens a new direction toward the physiological role of AcoP.
The electrochemical setup developed in this work allowed us to reconstitute on electrodes a portion of a respiratory chain and to quantify ET rates using small amounts of proteins and with the possibility to very easily tune the protein environment. We propose that this method could serve to get thermodynamic and kinetic data on other redox proteins, as well as between two interacting redox partners. Furthermore, for membrane proteins, the use of electrochemical interfaces appears more appropriate to mimic their physiological environment, in comparison with classical solution-state stopped-ow analysis.
Using the same strategy and by integrating the other partners of the respiratory chain, we should be able in the future to complete the understanding of the energy chain of A. ferrooxidans. For upstream Cyt c 4 , it was demonstrated that Rus was interacting with Cyt c 4 via Heme H . 7 Rus and AcoP are both cupredoxins of high redox potential. The purpose served by AcoP is not clear as Rus could be sufficient to shuttle electrons in the periplasm. Their stoichiometry (Rus is highly abundant) 55 and localization (membrane anchored for AcoP versus periplasm for Rus) might drive their specic function and explain their co-occurrence. It should be interesting to determine whether competition exists between the two cupredoxins in the respiratory chain, and under which conditions. Downstream, the question of the electron transfer to the terminal oxidase CcO Scheme 3 Proposed model of ET between redox proteins and the electrode surface observed on PG electrodes (see Fig. 8). Dashed lines represent PG electrodes. Structures of Cyt c 4 or predicted AcoP are presented in grey ribbons with their respective redox centers, hemes (in red) and copper (in blue). (A) Interfacial ET between the electrode and the two heme centers, a reversible process, illustrated by black arrows; (B) slow interfacial ET between the AcoP copper center and the electrode shown by a dashed arrow; (C) additional intermolecular ET pathway within the AcoP : Cyt c 4 complex depicted by a green arrow. This complex has been modelled with the HADDOCK web server 46 by driving the docking with the constraint based on AcoP interacting with Heme H of Cyt c 4 .
should be addressed. In most known biological systems, Cyt c has been described as the electron donor to the CcO. 56 However, in some biological systems, cupredoxins have been shown to replace Cyt c as electron donors to terminal enzymes depending on the metal occurrence in the bacterial environment. 57 In the case of A. ferrooxidans, AcoP and Cyt c 4 being under the control of the same promoter, and hence expressed under the same conditions, might both be able to transfer electrons to the CcO. We can hypothesize that AcoP may act as a second entry for electrons, while protecting the CcO copper center exposed to the acidic environment. 11 Further studies are required to answer such questions.
For a long time A. ferrooxidans has attracted the interest of industries regarding its use in bioleaching processes. More recently, its advantage as a biocathodic catalyst in microbial fuel cells 3 expanded the importance of its study, which emphasizes the need to better describe how this bacterium nds energy to grow using iron, a low energetic substrate. Even more interestingly, proteins in this ET chain share the common features of having a high redox potential. Such properties can be of great importance for ecologically friendly devices such as enzymatic fuel cells. These fuel cells are cathode-limited by the low oxygen affinity of currently used multicopper oxidases. CcO could be an alternative, but the most widely characterized CcOs display low redox potentials, a limitation that A. ferrooxidans CcO could overcome.

Conflicts of interest
There are no conicts to declare.