A biosynthetic model of cytochrome c oxidase as an electrocatalyst for oxygen reduction

Creating an artificial functional mimic of the mitochondrial enzyme cytochrome c oxidase (CcO) has been a long-term goal of the scientific community as such a mimic will not only add to our fundamental understanding of how CcO works but may also pave the way for efficient electrocatalysts for oxygen reduction in hydrogen/oxygen fuel cells. Here we develop an electrocatalyst for reducing oxygen to water under ambient conditions. We use site-directed mutants of myoglobin, where both the distal Cu and the redox-active tyrosine residue present in CcO are modelled. In situ Raman spectroscopy shows that this catalyst features very fast electron transfer rates, facile oxygen binding and O–O bond lysis. An electron transfer shunt from the electrode circumvents the slow dissociation of a ferric hydroxide species, which slows down native CcO (bovine 500 s−1), allowing electrocatalytic oxygen reduction rates of 5,000 s−1 for these biosynthetic models.

M imicking the sophistication of naturally occurring enzymes has been a long-term goal of the scientific community. An artificial analogue that can perform equally well as its natural predecessor will not only provide deeper understanding of the native enzymes, but also enable the development of efficient artificial catalysts. For several decades now chemists have embarked on this daunting pursuit of emulating the efficiency and selectivity of naturally occurring enzymes and several important milestones have been achieved. Efforts from synthetic inorganic chemists have resulted in synthetic models of myoglobin (Mb), galactose oxidase, tyrosinase, cytochrome P450 and cytochrome c oxidase (CcO) [1][2][3][4][5] . Alternatively, there has been fervent pursuit of biochemical constructs inspired by natural metalloenzymes. A series of binuclear non-haem iron, cytochrome c, haem oxidases and iron-sulfur enzyme models have resulted from such efforts 4,6-12 . While none of the synthetic or biochemical models reported so far could match the reactivity exhibited by their natural counterparts, fundamental insights regarding the structurefunction correlations of several metalloenzymes have been gained in the process 7,[13][14][15] . In addition, key information about the secondary coordination sphere interactions present in the protein-active site, which play a dominating role in determining the electronic structure and reactivity of these metalloenzymes, have been identified 16,17 .
In a biosynthetic approach, stable naturally occurring proteins have been used as scaffolds for creating mimics of several metalloenzymes, such as hydrogenases which are involved in the reversible generation of H 2 from water, haem proteins participating in electron transfer and O 2 -binding, non-haem iron and copper enzymes active in small molecule activation, and even novel enzymes containing non-native cofactors [18][19][20][21][22][23][24] . For example, using this approach, biosynthetic models that structurally and functionally mimic CcO and nitric oxide reductase have been reported 7,25 . Despite decades of focused effort, however, biosynthetic models with catalytic efficiencies approaching those of the naturally occurring metalloenzymes have remained elusive [26][27][28] . In this report, we communicate a biosynthetic model of CcO bearing the distal Cu B and a tyrosine residue that is kinetically more competent in reducing O 2 electrochemically than any known synthetic analogue, as well as native CcO itself.
X-ray crystallography of Mb and its mutant have revealed that its two propionate side chains project out of the protein surface into the solvent (Fig. 1) 7 . Taking advantage of this structural feature, we have previously developed an electrocatalytic O 2 reduction system where the native haem cofactor in Mb is replaced by a modified hemin cofactor bearing an alkyne group (Hemin-yne, Fig. 2) so that electrons can be injected directly into the haem from a gold electrode to facilitate O 2 reduction 29 . This method resulted in a Mb-functionalized electrode bearing 2.15 Â 10 À 12 mol per cm 2 of protein, which was characterized using several microscopic and spectroscopic techniques 29 . Over the last few years, a biosynthetic model of CcO has been reported in which two distal residues of Mb (L29 and F43) have been mutated to His, which along with the native His64, form a Cu-binding site, mimicking the distal Cu B -binding site present in CcO (Cu B Mb) 30 . Furthermore, in an attempt to mimic the conserved Tyr 244 residue in the CcO-active site, a G65Y mutant of Cu B Mb (G65YCu B Mb) containing redox-active tyrosine residue in the distal site and a variant where a tyrosine residue was crosslinked to the active site histidines were also created ( Fig. 1) 31 .
Herein we report the electrocatalytic properties of the G65YCu B Mb (higher synthetic yields than the tyrosine crosslinked variant) immobilized on an Au electrode using the method developed for WT Mb 29 .

Results
Electrode characterization by SERRS. Surface-enhanced resonance Raman spectroscopy (SERRS) data ( Fig. 3a) of the electrodes bearing the G65YCu B Mb protein with and without Cu B show the oxidation and spin state marker n 4 , n 3 , n 2 and n 10 bands at 1,375 cm À 1 , 1,493 cm À 1 ,1,585 cm À 1 and 1,641 cm À 1 , respectively. The n 4 , n 3 and n 2 values are consistent with the presence of a five-coordinated high spin haem in the active site on these electrodes bearing the biochemical constructs of CcO 32 . Also the bands at 1,504 cm À 1 and n 4 at 1,641 cm À 1 suggest the presence of a mixture of six-coordinate low spin species, which likely has H 2 O as the axial ligand. The positions of these bands in the G65YCu B Mb and their relative intensities are different from Hemin-yne ( Supplementary Fig. 1) 29 .
Electrode characterization by X-ray photoelectron spectroscopy. X-ray photoelectron spectroscopic (XPS) data of a G65YCu B Mbbound Au electrode clearly indicate the presence of Fe, Cu, C, N and O elements ( Supplementary Fig. 2 peak is broad ( Supplementary Fig. 2), as it contains several components due to the presence of amide, haem pyrroles and the triazole groups (resulting from the covalent attachment of Heminyne) on the surface 33 . Similarly, the C 1s peak (Supplementary Fig. 2) contains contributions from different types of C atoms (aromatic, aliphatic, haem and so on) on these protein-modified surfaces 36 .
Electrode characterization by CV. Cyclic voltammetry (CV) of G65YCu B Mb with and without the distal Cu B immobilized onto the electrodes in degassed buffer show the haem Fe 3 þ /2 þ midpoint reduction potential (E 1/2 ) at À 97 mV and À 57.5 mV, respectively (Fig. 3b). The peak separation between the cathodic and the anodic peak for both the cases is B70 mV (ref. 37). Hemin-yne displays the Fe 3 þ /2 þ reduction potential at À 70.0 mV in the absence of a protein and À 135.0 mV when bound to wild-type apo Mb 29 . In the case of the G65YCu B Mb protein-bound electrodes, the Cu 2 þ / þ process overlaps with the Hemin-yne Fe 3 þ /2 þ process, resulting in approximately twice the area under these CV peaks relative to the G65YCu B Mbbearing electrodes prior to Cu B loading. The integrated area under these CV features in the absence of Cu 2 þ indicates that there are 2.55 ± 0.05 Â 10 À 12 mol of protein per cm 2 of the surface. The ratios of the integrated area under the CV features of G65YCu B Mb functionalized before and after loading the Cu B is B1:2 (Table 1, fourth column) which is consistent with the expected 1:1 stoichiometry (that is, every G65YCu B Mb binds one Hemin-yne and one Cu 2 þ ion). Note that the E 1/2 values of the hemin and Cu B measured for these electrodes are slightly different from those estimated from potentiometric titration in solution 7,19 . This is likely due to the interfacial microenvironment of the -COOH-terminated Self Assembled Monolayer (SAM) which is known to shift the apparent formal potentials of redoxactive species in its vicinity 38 . Thus the in situ reconstitution of the protein with Hemin-yne on the electrode is evident from the SERRS, XPS ( Supplementary Fig. 2) and CV data ( Supplementary  Fig. 3, Supplementary Fig. 4). The presence of the Cu 2 þ at Cu B site on the electrode is indicated by XPS and CV data. Taken together, these data indicate the assembly of the G65YCu B Mb, biosynthetic model of CcO covalently attached to the electrode via the linkage between the Hemin-yne and the azide terminated thiols created using click reaction (Fig. 2). ARTICLE electrocatalytic O 2 reduction currents are observed by the G65YCu B Mb (with and without Cu 2 þ ) bearing bio-electrodes at pH 7 at room temperature, as the applied potential is lowered below þ 100 mV versus NHE (Fig. 4a, Supplementary Fig. 5). Thus as the potential of the electrode is lowered such that the iron in these proteins is reduced to Fe II , an electrocatalytic O 2 reduction current is observed. It is important to note that the potential of O 2 reduction reaction (ORR) (E ORR ) is À 263 mV, which is more negative than the E 1/2 ( À 97 mV), suggesting that the potential determining step of ORR is not the reduction of resting Fe III to Fe II but the reduction of a different species with À 166 mV more negative potential (Fig. 4a inset). A more negative E ORR relative to E 1/2 is mechanistically significant (vide infra).
In these active sites, O 2 may be reduced by 4e À and 4H þ to H 2 O, or by fewer electrons to produce partially reduced oxygen species (PROS) like O 2 À and H 2 O 2 . The extent of 4e À reduction and the second order rate constant (k ORR ) of the ORR can be determined using rotating disc electrochemistry (RDE) where the catalytic O 2 reduction current increases with increasing rotation rates ( Fig. 4a,b) following the Kouteky-Levich (K-L) equation (equation (1)) 39 .
where, i K (E) is the potential-dependent kinetic current and i L is the Levich current. i L is expressed as where n is the number of electrons transferred to the substrate, A is the macroscopic area of the disc (0.096 cm 2 ), [O 2 ] is the concentration of O 2 in an air-saturated buffer (0.26 mM) at 25°C,    bioelectrode at À 300 mV potential and at multiple rotation rates, with the inverse square root of the angular rotation rate (o À 1/2 ) (c) Difference between the potential for O 2 reduction (E ORR ) and the midpoint reduction potential of Fe 3 þ /2 þ redox couple (E 1/2 ). D O2 is the diffusion coefficient of O 2 (1.8 Â 10 À 5 cm 2 s À 1 ) at 25°C, o is the angular velocity of the disc and n is the kinematic viscosity of the solution (0.009 cm 2 s À 1 ) at 25°C (ref. 40). Plot of i cat À 1 at multiple rotation rates with the inverse square root of the angular rotation rate (i cat À 1 ) for G65YCu B Mb (with Cu 2 þ ) (Fig. 4b) is linear. The slope of K-L plot is expressed as 1/[n{0.62FA(D O2 ) 2/3 n À 1/6 }], which can be used to experimentally estimate the value of n where n is the number of electrons donated to the substrate, that is, O 2 . The slope obtained from the experimental data for G65YCu B Mb (Fig. 4b) is close to the theoretical slope (Fig. 4b, dotted purple line) expected for a 4e À process and very different from the slope for a 2e À process (Fig. 4b, dotted green line). Thus the G65YCu B Mb bioelectrode predominantly catalyses a 4e À /4H þ reduction of O 2 to H 2 O at pH 7.
The intercept of the K-L plot is the inverse of the kinetic where, n is the number of electrons, A is the geometric surface area, [O 2 ] is the bulk concentration of O 2 , G cat is the surface coverage of the catalyst (obtained from the integration of the anaerobic CV data) and k ORR is the second order rate constant for O 2 reduction estimated at À 300 mV. At this potential, in an oxygenated buffer, the G65YCu B Mb catalyst is involved in substrate diffusion-limited ORR. Using this equation (equation (3)) and the experimentally obtained i K (E) at À 300 mV, the second order rate constant for O 2 reduction G65YCu B Mb is evaluated to be 1.98 Â 10 7 M À 1 s À 1 ( Table 2). The pseudo first order rate can be determined (  Fig. 6). The catalytic ORR rate by G65YCu B Mb surpasses those reported for the best artificial synthetic analogues ( Table 2). The G65YCu B Mb biosynthetic Mb scaffold-based bio-electrode for O 2 reduction is remarkably stable. Monolayers bearing covalently attached O 2 -reducing electrocatalysts reported so far have never been stable enough to allow these dynamic electrochemical experiments to determine the kinetic parameters (k ORR , number of electrons and so on). Enzymes like laccases, directly attached to chemically modified graphite electrodes, were found to be stable enough to be investigated with these hydrodynamic techniques 27 . The failure to perform these experiments has been attributed to degradation of the catalyst during RDE experiment, that is, very small turnover numbers presumably due to the production of PROS during ORR. Rotating ring disc electrochemistry (RRDE) shows formation of only B6% PROS by the G65YCu B Mb (Supplementary Fig. 7) during ORR, indicating that it reduces 96% of O 2 to H 2 O consistent with the RDE data. During the RDE experiments (Fig. 4a), the G65YCu B Mb-functionalized electrodes bearing 10 À 12 mol of the catalyst reduced 1.8 ± 0.3 Â 10 À 8 mol of O 2 (7 ± 1 Â 10 À 3 C total charge and 4e À per O 2 molecule) yielding a turnover number of at least 10 4 . The role of PROS in degrading the catalyst is established by the fact that the electrolytic current (at À 0.3 V) remains stable in the presence of 50 mM catalase in solution ( Supplementary Fig. 8).

Discussion
To understand the facile and selective O 2 reduction catalysed by the G65YCu B Mb biochemical model, the recently developed SERRS-RDE technique is employed 42 . In this technique, the rR spectra of the catalyst (that is, G65YCu B Mb) bearing electrode is collected while the system is involved in steady state O 2 reduction and the species accumulated in the steady state can be identified. For any species to accumulate in steady state, its rate of formation has to be greater than its rate of decay. Thus, while the species preceding the rate-determining step (rds) will accumulate at steady state, the accumulation of a species in steady state does not immediately imply its decay as the rds. In the absence of O 2, a high spin ferrous species is formed, characterized by a n 4 and n 3 vibrations at 1,357 cm À 1 and 1,473 cm À 1 (Fig. 5a, cyan), respectively, when a cathodic potential of À 0.4 V is applied signifying reduction of the resting ferric state (Fig. 5a, red) to the active ferrous state at these potentials. However, when the same reducing potential is applied in an oxygenated buffer the SERRS-RDE data ( Supplementary Fig. 9) clearly show the presence of different species during electrocatalytic ORR, which leads to broadening of the n 4 , n 3 and n 2 regions (Fig. 5a, green) relative to the oxidized and reduced states (Fig. 5a, red and cyan). In particular, the n 3 and n 2 vibrations discernibly shift to higher energies as indicated by clear increase in intensities at 1,508 cm À 1 and 1,591 cm À 1 (Fig. 4b), suggesting the accumulation of Fe IV ¼ O species during steady state ORR 43 . Signals from high spin ferrous, resting high spin ferric, low spin ferric and ferryl species with n 3 at 1,473 cm À 1 (Fig. 5a, cyan; and Fig. 5b, brown), 1,493 cm À 1 (Fig. 5a, red and Fig. 5b, dashed green), 1,504 cm À 1 (Fig. 5a, cyan and Fig. 5b, cyan) and 1,508 cm À 1 (Fig. 5a, green and Fig. 5b, green), respectively, could be convoluted by fitting the n 3 region of the spectrum. The lack of significant signal from the high spin ferrous species (weak n 3 at 1,473 cm À 1 ) suggests that O 2 binding to these species is facile in the steady state. These mutants use the basic design of Mb which has a very fast O 2 -binding rate (10 7 M À 1 s À 1 ) (refs 44,45). This rate is indeed B10 times faster than O 2 binding to the haem a 3 site of CcO (ref. 46). Similarly, the very weak intensity of the high spin (HS) Fe III species indicates that the ET to Fe III resting state is very facile at these potentials as may be expected due to direct attachment of the Hemin-yne to the electrode. The significant intensity of ferryl species (n 2 at 1,591 cm À 1 and n 3 at 1,508 cm À 1 ) entails the O-O bond cleavage leading to its formation to be faster than its decay via reduction under steady state. Thus the reduction of the resting Fe III state, O 2 binding to Fe II are facile in G65YCu B Mb under the reaction conditions. The low spin ferric species accumulated during steady state ORR could be a dioxygen adduct or peroxide adduct similar to those observed in native CcO and its model systems [46][47][48][49][50][51] . The low frequency region shows 18 O 2 -sensitive bands suggestive of the formation of a low spin ferric peroxide and Fe(IV) ¼ O (Supplementary Fig. 10).
If one were to conceive of a Gedanken steady state turnover experiment with CcO where the electron transfer to the active site is very fast (that is, in the hypothetical situation where ET from Cyt c to CcO is not the rds) as the ET from the electrode to the active site is very fast due to direct attachment of the later to the former, the species that would accumulate during turnover, based on the Babkock-Wikström mechanism (Fig. 6, the parameters of native CcO are indicated in purple and G65YCu B Mb are and Fe III -OOH species will have Raman signatures of low spin haem (Fig. 6), Fe III -OH will have rR signature of high spin haem and the Fe IV ¼ O will have signatures unique to haem ferryl species 46,54,55 . The SERRS-RDE data show the presence of species having signatures of low spin Fe III and Fe IV ¼ O. While the later can originate from only a single species, the former can indicate the presence of any of the three species or a combination of them. The lack of significant high spin signal indicates that the biosynthetic model circumvents accumulation of Fe III -OH and resting Fe III species in the steady state by facile ET. The overall rate-limiting step of native CcO in solution is the dissociation of hydroxide of the Fe III -OH end product of O 2 reduction from haem to generate the active ferric resting form and has a first order rate constant of 500 s À 1 (refs 46,54-56). This dissociation is required during turnover as the potential of this hydroxidebound form is likely to be more negative that the five-coordinate resting oxidized site (which will be regenerated after hydroxide dissociation) and will not be reduced by haem a. While the E 1/2 of a haem a 3 Fe III -OH species cannot be determined with confidence due to strong interaction potential and co-operativity between the haem a and haem a 3 sites, the potential of formate and azidebound high spin haem a 3 site (analogous to hydroxide) is B130 to 200 mV more negative than the resting ferric site 57,58 . The potential determining step of ORR (defined as the electron transfer (ET) step in catalysis having lowest potential) by the G65YCu B Mb is 166 mV more negative than the E 1/2 for the resting high spin ferric state and is likely to be the reduction of the Fe III -OH species. Thus the bioelectrode can circumvent the kinetic barrier associated with the dissociation of the hydroxide by directly reducing it to ferrous at 166 mV lower potential. This direct electron transfer to the ferric hydroxide species, which is an intermediate in the catalytic cycle of CcO, circumventing a slow step in catalysis, is an electron transfer shunt analogous to peroxide shunt in cytochrome P450, which overcomes the ratedetermining O 2 activation step of the native enzyme 59 . In the mass transfer controlled region of the catalytic current, all ET steps are facile and steps like O 2 binding, protonation and O-O bond cleavage can be rds at these potential. The lack of HS Fe II accumulation in the SERRS-RDE indicates that O 2 binding to Fe II is very fast and not the rds. The O-O bond cleavage in the Babkock-Wikström mechanism involves ET to the active site and not the rds as well. Thus rds of ORR by these biosynthetic models is likely to be the protonation of the Fe III -O 2 À species with a first order rate of 5,000 s À 1 (Fig. 6).
In an air-saturated buffer (0.26 mM O 2 ), the pseudo first order rate constant of ORR by G65YCu B Mb is determined to be B5-6 Â 10 3 s À 1 (k ORR [O 2 ]).The highest second order O 2 reduction rate reported for any synthetic mimic of CcO is 1.2 Â 10 5 M À 1 s À 1 ; that too on a multilayer having 1,000 times more catalyst than the G65YCu B Mb electrodes 60   constant of G65YCu B Mb is 10 7 M À 1 s À 1 which is, thus, 2 orders of magnitude higher than best synthetic haem/Cu-based O 2 reduction electrocatalyst. Thus the selectivity and kinetic rate of the G65YCu B Mb-bearing electrode surpasses those reported for smaller synthetic analogues and illustrates the advantages of using a biochemical scaffold over a synthetic scaffold. Although the pseudo first order rate constant of the G65YCu B Mb is 10 times faster than the rate of native CcO in solution, such a comparison is vulnerable to differences in reaction conditions (for example, G65YCu B Mb is water soluble but CcO exists in membranes). Alternatively, erstwhile efforts resulting in electrodes bearing native CcO in a manner similar to these bio-electrodes show extremely sluggish O 2 reduction 60-62 . This is due to improper alignment of this membrane-bound protein on the electrode, which precludes efficient electron transfer to the active site [63][64][65] . However, the direct attachment of haem to the electrode utilizing its solvent-exposed propionate groups (that is, a short circuit) enables fast electron transfer to the active site 29 . This is further supported by the fact that when ethynylferrocene (Fc) is attached to the same surface B25 mV peak separation is observed even at 5 V s À 1 (Supplementary Fig. 11) 38 , suggesting that the ET is indeed fast. As a result when a CcO-functionalized SAM-covered Au electrode produces o1 mA electrochemical O 2 reduction current at À 300 mV, this bio-electrode produce B100 mA current at similar potentials 66 .
Finally, the G65YCu B Mb mutant has residues in the distal pocket that can help both electron and proton transfer during O 2 reduction (Y65 in G65YCu B Mb). In CcO, the involvement of Tyr 244 residues in proton/electron transfer during O 2 reduction is now widely accepted 67,68 . Previous biochemical and structural studies on these mutants had indeed indicated the close proximity of this residue to the distal site 7,19,30 . An analogous biochemical model without the Y65 residue, Cu B Mb, is not as stable as the G65YCu B Mb as the former degrades rapidly during the RDE experiments ( Supplementary Fig. 12). In summary, a electron transfer shunt which circumvents the rate-determining dissociation of a ferric hydroxide species by directly reducing it at slightly negative potential, fast O 2 binding, fast electron transfer to the active site and the presence of a protective Y65 residue in a biochemical model of CcO results in O 2 reduction activity 100 times faster than the best synthetic models, order of magnitude faster than CcO immobilized on electrode and follows a mechanism comparable to that of native CcO in solution.

Methods
Materials. 1-Azidoundecane-11-thiol and Hemin-yne were synthesized following the reported procedure 29,69 . 6-Mercaptohexanoic acid was purchased from Sigma Aldrich. Di-sodium hydrogenphosphate dihydrate (Na 2 HPO 4 . 2H 2 O) was purchased from Merck. 2, 6-lutidine was purchased from Avra Synthesis Pvt. Ltd. These chemicals were used without further purification. Au wafers were purchased from Platypus Technologies (1,000 Å of Au on 50 Å of Ti adhesion layer on top of a Si(III) surface). Transparent Au wafers (100 Å of Au on 10 Å of Ti) were purchased from Phasis, Switzerland. Au and Ag discs for the RRDE and SERRS experiments, respectively, were purchased from Pine Instruments, USA. The Mb mutants were prepared as reported in the literature 7,30 . Analysis of the components of the rR spectrum was done by using Lorenztian line shape of peak fit software.
Instrumentation. All electrochemical experiments were performed using a CH Instruments (model CHI710D Electrochemical Analyzer). Bipotentiostat, reference electrode and Teflon plate material evaluating cell (ALS, Japan; http://www.als-japan.com/1398.html) were purchased from CH Instruments. The RRDE set-up from Pine Research Instrumentation (E6 series ChangeDisk tips with AFE6M rotor) was used to obtain the RRDE data. The mutant Mb-functionalized or SAM-covered Au surface (disc of 0.1 cm 2 area for RDE, RRDE and wafer of 0.45 cm 2 area for CV) was always used as the working electrode. The XPS data were collected in a Omicron (model: 1712-62-11) spectrometer using a high-resolution monochromatic Al-Ka source at 1,486.7 eV under 15 kV voltage and 10 mA current maintaining a base pressure of 5 Â 10 À 10 mbar. The binding energies were calibrated to the Ag 3d 5/2 peak at 368.2 eV. The resonance Raman experiments were done in the Kr þ Laser (Sabre Innova, Model-SBRC-DBW-K) purchased from Coherent, and the data were collected using the Spectrograph (Model-Trivista 555) from Princeton Instruments.
Formation of mixed SAM and covalent attachment of Hemin-yne to it. Mixed self-assembled monolayer of 1-azidoundecan-11-thiol and 6-mercaptohexanoic acid was formed on immersing the properly cleaned Au wafers or disks into the deposition solution containing 1-azidoundecan-11-thiol and 6-mercaptohexanoic acid in 10 ml of ethanol in a desired ratio (typically 1:49). The total thiol concentration of these deposition solutions were always maintained at 1 mM. On this SAM Hemin-yne was covalently attached using 'Click' reaction 29 .
Reconstitution of Apo-G65YCu B Mb mutants to G65YCu B Mb. For all the experiments on heterogeneous SAM surfaces, the Hemin-yne modified -COOH SAM surfaces were incubated with a 20 mM apoprotein (Apo-G65YCu B Mb) solution for 2 h. The supernatant solution was drained and the surface was cleaned with water. The presence of Cu 2 þ in the non-haem-binding site is confirmed by electron paramagnetic resonance (EPR) spectroscopy ( Supplementary Fig. 13). The immobilization of the mutant is further confirmed by the absorption spectra and SERRS of the surface fabricated with those mutants (Supplementary Fig. 14,  Supplementary Fig. 15).
Cyclic voltammetry. The CV was performed using Au wafers sandwiched between two Teflon blocks of the Plate material evaluating cell. All electrochemical experiments were done in pH 7 phosphate buffer containing potassium hexafluorophosphate. Anaerobic cyclic voltammetric experiments were done by using degassed buffer (three cycles of freeze-pump-thaw). Ag/AgCl reference electrode and Pt counter electrode were used throughout all the electrochemical experiments except the case of anaerobic experiments where only Ag wire was used as the reference electrode.
The peak areas were estimated by integrating the anodic/cathodic peak of the anaerobic CV of the mutant Mb-functionalized SAM-covered Au surface, using the data acquisition software itself. A line collinear with the background is used to subtract the background. The estimated area has been further confirmed by subtracting the background current of a SAM-functionalized electrode ( Supplementary Fig. 11) bearing ferrocene. Both these approaches provide the same estimate.
To ensure that the SAM surface is stable during the electrocatalytic investigations, disc bearing just the SAM was subjected to several rotations (200-1,000 r.p.m.) and its capacitive current was found not to change, indicating that the SAM is retained on the electrode during these dynamic electrochemistry experiments ( Supplementary Fig. 16). SAM can also be damaged when the protein atop the SAM degrades during ORR due to the reactive oxygen species produced. When an unstable electrocatalyst (Hemin-yne) decayed there was a steady loss of ORR current, indicating degradation of the active site but the capacitive currents of the SAM were unaltered (Supplementary Fig. 17).
SERRS and SERRS-RDE. The excitation wavelength used in the Resonance Raman experiments was 406.7 nm and the power applied to the sample was 10-15 mW. The spectrograph was calibrated against naphthalene. The Ag surfaces were roughened before SERRS experiments following literature protocols 70 . The SERRS-RDE set-up is described in ref. 42. The data for the oxidized state was obtained by holding the potential of the disc at 0 mV versus NHE, and the data during steady state ORR was obtained by holding the disc at À 400 mV versus NHE and the disc was rotated at 300 r.p.m. Normally data were acquired over a period of 300 s.