Supporting Information

Supporting Information

Direct Measurement of Charge Regulation in Metalloprotein Electron Transfer Collin T. Zahler, Hongyu Zhou, Alireza Abdolvahabi, Rebecca L. Holden, Sanaz Rasouli, Peng Tao, and Bryan F. Shaw* anie_201712306_sm_miscellaneous_information.pdf

Supporting Materials and Methods: Synthesis of protein charge ladders: Protein charge ladders were synthesized by acetylating lysine--NH3+ with acetic anhydride, as previously described.[1] For myoglobin (Mb) ladders, we synthesized charge ladders of oxidized Fe3+-Mb (met-Mb) because commercially available Mb (lyophilized powder from Sigma-Aldrich) is in the oxidized form due to prolonged exposure to O2. Briefly, Fe3+-Mb was dissolved in 100 mM HEPBS buffer pH 9.0, to a final concentration of 160 μM. Partial acetylation of Fe3+-Mb was achieved via addition of ~ 5 molar equivalents of acetic anhydride (dissolved in 1,4-dioxane) to Fe3+-Mb solution. Acetylated Fe3+-Mb solutions were then transferred to 10 mM KPO4 buffer (pH 7.4) using centrifugal filtration (4000 × g, 4 C). Charge ladders of Cu2+-Azurin (Az) and Fe2+-Cytochrome c (Cyt c) were synthesized using the same method, except that acetylated Fe2+-Cyt c was transferred to tris-glycine buffer (25 mM tris, 192 mM glycine, pH 8.3). Charge ladders for Fe2+-Mb (deoxy-Mb) and Cu1+-Az were generated via reducing the acetylated Fe3+-Mb samples (or Cu2+-Az) with sodium dithionite (Na2S2O4). A 1 M solution of Na2S2O4 was prepared in degassed 10 mM KPO4 buffer (pH 7.4), 100 molar equivalents of which was added to each protein solution immediately prior to each CE experiment. Reduction was confirmed using UV-vis spectrophotometry both before and after CE. All solutions were deoxygenated via N2 bubbling before analysis with CE. Charge ladders for Fe3+-Cyt c were generated via oxidizing the acetylated Fe2+-Cyt c with potassium ferricyanide (K3[Fe(CN)6]). A 1 M K3[Fe(CN)6] solution was prepared in tris-glycine buffer, and ~ 5 molar equivalents were added to Fe2+-Cyt c charge ladder. Oxidation was confirmed via UV-vis spectrophotometry both before and after analysis with CE.

2

Capillary Electrophoresis (CE): All CE experiments were performed using a Beckman P/ACE MDQ, equipped with a bare fused-silica capillary. The capillary was enclosed in a liquid cooled jacket to prevent Joule heating. One microliter of 1 M dimethylformamide (DMF) was added to each sample prior to electrophoresis as a neutral marker of electroosmotic flow (EOF). Values of migration were expressed in units of mobility (cm2·kV-1·min-1), and normalized to the migration of the neutral marker (this corrects for any variations in electroosmotic flow among iterate runs). For Mb experiments, electrophoresis was performed at 25 kV (22 C), using deoxygenated 10 mM KPO4 (pH 7.4) as the running buffer. The total run time for each CE experiment was 15 minutes. The final concentration of each Mb solution analyzed by CE (both Fe3+-Mb and Fe2+Mb) was ~ 100 μM, determined using UV-vis spectrophotometry (ε280nm = 13,940 M-1cm-1). For Fe2+-Mb experiments, 100 molar equivalents of Na2S2O4 were added immediately prior to each CE experiment. UV-vis spectrophotometry was performed on Fe2+-Mb charge ladders immediately after each CE experiment to ensure the absence of oxidation (Figure S3 b). For Az experiments, electrophoresis was performed using the same method as Mb, except that the voltage was increased to 29 kV (electrophoresis persisted for ~10 minutes). Concentration of Az was confirmed with UV-vis (Cu2+-Az; ε626nm = 4,800 M-1cm-1). Cu1+-Az charge ladders were generated by reduction with 100 molar equivalents of Na2S2O4. Cu1+-Az remained stable (i.e., did not re-oxidize) throughout the course of all experiments, verified with UV-vis before and after CE analysis (Figure S3 f) For Cyt c experiments, electrophoresis was performed at 29 kV (22 °C), using deoxygenated tris-glycine as the running buffer. We used tris-glycine buffer instead of 10 mM KPO4 (pH 7.4) as running buffer for CE experiments of Cyt c because the electroosmotic flow in 10 mM KPO4 was found to be too slow in the presence of polybrene-coated capillary, which

3

hindered the accurate determination of mobility, especially for rungs with higher mobilities (i.e., Ac(1) and Ac(0)). The relative mobility of rungs of Cyt c are reversed compared to Mb and Az because Cyt c is a net positively charged protein (unlike Mb and Az). Thus, Cyt c cannot be analyzed with a bare fused silica capillary because the protein adheres to the anionic silica surface. To prevent this electrostatic interaction, the capillary was coated with 7.5 %, positivelycharged polybrene, as described previously.[2] The final concentration of Cyt c for each CE experiment was ~ 230 μM, determined using UV-vis spectrophotometry (ε280nm = 11,580 M-1cm1

). For Fe3+-Cyt c experiments, 5 molar equivalents of K3[Fe(CN)6] were added to each protein

solution immediately prior to each electrophoresis experiment. Fe3+-Cyt c was significantly more stable than Fe2+-Mb, and thus multiple electropherograms could be obtained from each sample (i.e., the oxidation state was verified after one, five and ten CE experiments, using UV-vis, Figure S3 d). In contrast, replicate runs of Az and Mb required freshly reduced protein.

Determination of protein net charge using CE: The migration time of unacetylated (Ac(0)) protein and each rung of the charge ladder (Ac(N)) was converted to electrophoretic mobility, using Equation (1):



ltot ld 1 1 (  ) V t t DMF

(Equation 1)

Where ltot is the total length of the capillary (60.2 cm), ld is the length from the beginning of the capillary to the detector (50.0 cm), V is the applied voltage (Mb: 25 kV; Cyt c, Az: 29 kV), t is the migration time of proteins, and tDMF is the migration time of the neutral marker (DMF). The mobility of protein species (μ), is expressed in units of cm2·kV-1·min-1. To calculate the net charge (ZMb, ZCyt c, ZAz), values of mobility of each rung of the electropherogram (μN) for each protein were plotted against their corresponding acetylation 4

number (N). A linear function was fit to the data for each set of replicate electropherograms according to Equation (2):

N 

e( Z Mb  N Z ) f eff

(Equation 2)

Where ΔZ is the change in charge upon acetylation of lysines in each protein, which is ~ -0.9 units instead of -1.0 because of charge regulation,[3] feff is the hydrodynamic drag, and e is the charge of electron.

Electrospray ionization mass spectrometry (ESI-MS): A Waters Synapt G2S ESI/TOF mass spectrometer was used to measure the extent of acetylation (Figure S1). Acetylated protein solutions were diluted with 0.1 % formic acid to a final concentration of 1 μM, loaded on a C-18 reversed-phase trapping column, and eluted with a 40:60 mixture of 0.1 % formic acid in water and pure acetonitrile for 6 minutes. Mass spectra were deconvoluted using MaxEnt1 module in MassLynx software.

Trypsin digestion and tandem mass spectrometry: Tryptic digests of partially acetylated protein solutions were generated via incubation with (i) trypsin Gold (Promega®, WI, USA) at a ratio of 1:20 (trypsin:acetylated protein), and (ii) dithiothreitol (DTT; 2 mM) to ensure the complete reduction of cysteine residues. Digestion of acetylated proteins was performed in 50 mM Tris-HCl buffer, pH 8.8. The final concentration of protein for digestion was 150 μM. The tryptic digest mixture was incubated at 37 C for 24 h, prior to the analysis with mass spectrometry. A LTQ LX/Orbitrap Discovery LC/MS instrument (Thermo Scientific™) was used to acquire mass spectra for tryptic peptides (Figure S2). The proteomics analysis was

5

performed using the SEQUEST software. The following criteria were set for maximizing the accuracy of proteomics analysis: mass difference tolerance < 5 ppm, Xcorr > 6.

Calculation of pKa values for ionizable amino acid residues: The x-ray crystal structures for oxidized and reduced states of each protein were obtained from the Protein Data Bank (PDB): for Mb, 1YMB (Fe3+-Mb) and 5D5R (Fe2+-Mb), for Cyt c, 1AKK (Fe3+-Cyt c) and 1GIW (Fe2+Cyt c), and for Az, 5AZU (Cu2+-Az) and 1E5Z (Cu1+-Az). To prepare the simulation system from these crystal structures, missing hydrogen atoms were added to the protein structures using the CHARMM program.[4] CHARMM force field version 22[5] was applied on proteins in this study. Iron and copper ions in the oxidized and reduced state in these proteins were treated explicitly with their corresponding charges and force field parameters. The partial charges and related force field parameters were adopted from previous computational studies for Fe3+ and Fe2+ heme,[6] and Cu2+ and Cu1+.[7] For Mb, partial charges of H93 and H2O-156 were also adjusted to maintain charge balance.[6] Both reduced and oxidized proteins were immersed into water to build a cubic simulation box using the TIP3P explicit water model.[8] The size of the simulation boxes for reduced and oxidized Mb, Cyt c, and Az were 66.58 Å and 66.35 Å, 57.80 Å and 56.20 Å, 60.47 Å and 60.49 Å, respectively. Sodium and chloride ions were added to simulation boxes to achieve net zero charge and maintain 0.1 M ionic strength. After minimization and heating processes, Langevin molecular dynamics simulations were performed for both proteins at a constant temperature (300 K) and volume. For each system, three independent simulations of 16 ns were carried out. The first 6 ns of the simulation were treated at equilibrium, and only the last 10 ns were used for pKa calculation, which provides a total sampling time of 30 ns for each system. Snapshots with the interval 0.1 ns were used to calculate

6

the pKa shift for each residue using the linear Poisson-Boltzmann equation implemented in CHARMM.[9] The optimized sets of Born-like radii (which were determined by molecular dynamics simulations with explicit water model in CHARMM force field),[5] and the corresponding van der Waal dielectric boundary condition, were used for all calculations.[10] The solvent dielectric used in calculations was  = 80, and the protein dielectric used in calculations was  = 20, however, we also varied the protein dielectric from 1 to 80 to investigate the effect of protein dielectric on computed values; see supporting results and discussion. Final pKa values for each residue were calculated by averaging all snapshots of each residue’s ΔpKa, as previously described.[11]

Supporting Results and Discussion: Determination of metal oxidation state using UV-vis: The oxidation state of each metalloprotein was verified using UV-vis spectrophotometry based on characteristic absorbance maxima for each protein species: Fe3+-Mb: Soret band = 408 nm, Q band = 502 nm; Fe2+-Mb (deoxy-Mb): Soret band = 433 nm, Q band = 560 nm; Fe3+-Cyt c: Soret band = 408 nm, Q band = 530 nm; Fe2+-Cyt c: Soret band = 413 nm, Q1 band = 519 nm, Q2 band = 549 nm; Cu2+-Az: Soret band = 407 nm, d-d band = 626 nm; Cu1+-Az: Soret band = 415 nm, β band = 520 nm, α band = 550 nm (Figure S3).[12-13] The decrease in intensity of the Soret band of Fe2+-Mb compared to Fe3+-Mb that we observed (a reduction from 0.78 a.u. to 0.50 a.u.; Figure S3 a) confirms the total reduction of Fe3+-Mb to Fe2+-Mb and is consistent with reported differences in the molar extinction coefficient of Fe3+-Mb (i.e., 188 mM-1cm-1) and Fe2+-Mb (i.e., 121 mM-1cm1 [14]

).

Observation of a single isosbestic point in the UV-vis spectra of Fe3+-Mb and Fe2+-Mb (at

418 nm; Figure S3 a), Fe3+-Cyt c and Fe2+-Cyt c (at 410 nm; Figure S3 b), Cu2+-Az and Cu1+-Az 7

(at 409 nm; Figure S3 c) confirmed the absence of other species, e.g., partially oxidized or partially hydrated proteins.

The oxidation state of each metal center does not change during the course of electrophoresis: In order to ensure that the redox state did not change throughout the course of each CE run (Figure 1 b-g and S4), we analyzed each of the iterate solutions of Fe2+-Mb, Fe3+Cyt c, and Cu1+-Az with UV-vis at the end of each replicate CE experiment (Figure S3 c-e). UVvis confirmed that Fe2+-Mb and Cu1+-Az proteins remained reduced, and Fe3+-Cyt c proteins remained oxidized over the brief period of electrophoresis. CE was initiated < 2 min after reduction/oxidation, and electrophoresis is completed in < 15 min, however, the first two rungs of the protein charge ladders pass through the UV-vis detector of the CE instrument after ~ 2 min. The electrophoresis of protein charge ladders was performed in running buffers that were deoxygenated via N2 bubbling for > 2 hours and used immediately for CE.

Addition of reducing/oxidizing agents causes peak broadening during CE: The presence of 10 mM Na2S2O4 in the Fe2+-Mb and Cu1+-Az samples (or 1 mM K3[Fe(CN)6] in the solution of Fe3+-Cyt c) caused the broadening of DMF marker and all replicate rungs of charge ladders (Figure 1 b-g and S4-5). This broadening is most likely an artifact of electrophoresis, because it was observed in the DMF peak. Peak broadening is a common (albeit enigmatic) artifact in CE, with multiple causes that are not mutually exclusive, e.g., fluctuations in EOF, electrokinetic injection, or electromigration dispersion.[15-16] The causes of broadening here are not known, however, reductions in EOF would cause a fraction of observed broadening: Na2S2O4 changed EOF by ~ 1 min, and K3[Fe(CN)6] changed EOF by ~ 1.2 min (Figure S5).

8

Prediction of the mechanism of charge regulation during single electron transfer (ET) using Poisson-Boltzmann electrostatic calculations: We note that our calculations of pKa values of all ionizable residues in Mb also contained the previously measured pKa of H2O[17] that directly coordinates to iron in Fe3+-Mb and dissociates upon reduction. A list of all calculated pKa values for all proteins (at both redox states) can be found in Tables S1-S6. According to the calculated pKa values for Mb (Tables S1-S2), the reduction of Fe3+-Mb with Na2S2O4 results in substantial changes in charge in eleven ionizable amino acid residues, and the water molecule coordinating iron (“substantial” is arbitrarily defined in this paper as ΔZ ≥ 0.01). These residues include all ten non-coordinating histidine residues (H24, H36, H48, H64, H81, H82, H97, H113, H116, and H119), α-NH3-G1, and the coordinating H2O in Fe3+-Mb (Figure S6 and Table S1-S2). The largest changes in charge following reduction involved: (i) a decrease in pKa of H81 by 0.66 units (ΔZH81 = -0.36); (ii) a decrease in pKa of H24 by 0.66 units (ΔZH24 = -0.35; Figure S6 and Tables S1-S2); and (iii) an increase in pKa of H97 by 0.56 units (ΔZH97 = +0.20; Figure S6 and Tables S1-S2). Calculated differences in charge (pKa) of many other residues in Fe2+-Mb and Fe3+-Mb are small (i.e., ΔZ < 0.01, Tables S1-S2) and are excluded (for clarity) from the graphical mechanism of charge regulation in Figure S6 (but are not excluded from calculations). Our mechanism of charge regulation in Cyt c is similar in nature to that of Mb. The oxidation of Fe2+-Cyt c with K3[Fe(CN)6] leads to significant changes in the charge of amino acid residues (H33; ΔZH33 = +0.03, K13; ΔZK13 = +0.01, K87; ΔZK87 = -0.01, Y48; ΔZY48 = +0.02, Y97; ΔZY97 = +0.01, and α-NH3-G1; ΔZNT-G1 = +0.07; Figure S7 and Tables S3-S4). The calculated differences in the charge (pKa) of all residues in Fe2+-Cyt c and Fe3+-Cyt c are shown

9

in Tables S3-S4, and are excluded from the graphical mechanism (but not from calculations) of charge regulation in Figure S7. Single ET from Na2S2O4 to Cu2+-Az only leads to significant changes in the pKa of three residues, the two non-coordinating histidines (ΔZH35 = +0.14; ΔZH83 = +0.16) and α-NH3-A1 (ΔZNT-A1 = +0.02). The calculated differences in the charge (pKa) of all residues in Cu2+-Az and Cu1+-Az are shown in Tables S5-S6, and are excluded for clarity from the graphical representation (but not from calculations) of charge regulation in Figure S8. Are these mechanisms reasonable, even though they would require long-range electrostatic perturbations to residues that are distal to the active site (in all three proteins), while excluding electrostatic interactions with the proximal histidine residues? For example, H81 in Mb is 23.4 Å from the heme-Fe center (Figure S6). Why would distant residues such as H81 exhibit such significant changes in charge (and pKa) upon ET, i.e., interact electrostatically (presumably) with Fe+2/+3, while ionizable functional groups closest to Fe-heme do not appear to undergo large changes in charge and pKa? First and foremost, H81 (and all other implicated residues) is close enough to Fe-heme to interact with it electrostatically (Figure S6-S8). The Debye length (κ-1) in our buffer is κ-1 ≈ 21 Å at 25 °C (assuming εwater = 78; I = 20 mM), however, the Debye length through the hydrophobic core of the protein is greater, so much that charged residues interact at dimensions similar to the size of the protein (e.g., κ-1 > 70 Å, assuming εprotein = 4 and I < 100 μM in the protein interior). Therefore, the distal residues are by no means too far from metal center to be affected (electrostatically) by its reduction/oxidation (Figure S6-S8). Secondly, residues that undergo the greatest change in charge in response to a change in electrostatic environment are—by definition—residues with pKa values nearest the pH of the

10

local environment (Figure S9). This is why we suspect that “proximal” histidines are not likely involved in charge regulation. For example, the pKa of proximal histidine in Mb (i.e., H93) that directly coordinates Fe2+/3+ is so low (at pH 7.4) that H93 will not undergo a significant change in charge (the pKa of histidine residues coordinating Fe2+/3+ are difficult to determine,[18-19] however, the pKa of H93 has been inferred from ligand dissociation studies[20] to be pKa ~ 3.5). Perhaps the strongest piece of evidence for redox-coupled protonation of ionizable residues that involves non-allosteric, direct electrostatic interactions is the classic study by Varadarajan et al. that reports the observation of the V68E variant of human Mb wherein E68 is deprotonated in Fe3+-Mb, but becomes protonated upon reduction to Fe2+-Mb.[21] This study has also shown that the insertion of ionizable residues in the hydrophobic core of human Mb (e.g., V68D and V68E) reduces the redox potential of Mb from Eº' ≈ +60 mV to ≈ -135 mV (whereas Eº' ≈ -24 mV for V68N).[21] These differences in potential were attributed to redox-induced protonation of D68 and E68 in Fe2+-Mb (i.e., uptake of one H+ by D68 or E68 upon reduction of Fe3+ to Fe2+).[21] The dielectric constant for proteins is generally estimated to range from 4 to 7 at the hydrophobic interior to 20 to 30 at surface residues of a protein.[22-23] To determine the robustness of Poisson-Boltzmann calculations of pKa (and ΔΔGZ and ΔZ) as a function of protein dielectric constant, we ran calculations at ε = 1 to 80. Plots of computed ΔG (of each redox state), ΔΔGZ and ΔZ at ε = 1 to 80 show that values remain relatively stable at ε = 4 to 80 (Figure S11 a-d). The values of Z are stable for Cyt c and Az across  = 4 to 20, and vary by 6 % and 29 %. For Mb, Z decreases 2.5-fold from  = 4 to 20. From  = 4 to 20, ΔΔGZ varies 3-fold for Cyt c, and 10 to 28 % for Mb and Az, Figure S11 b). We note that values of ΔΔGZ for Mb mirror previously calculated trends in reorganization energy at low .[24] For all proteins, as 

11

approaches zero, the variation in G and ΔΔGZ (of each 0.1 ns snapshot) increases, and the reliability of the average values becomes less significant at  < 4 (Figure S11 b-d). For the theoretical values of ΔZ and ΔΔGZ reported in the main text (Table 1, Figures 3 and S9-10), we used a protein dielectric constant of ε = 20. We hypothesize that calculations at ε = 20 are most valid because we are interested in examining changes in protonation of ionizable surface residues via movement of protons to and from solvent.

Supporting Figures:

Figure S1. Electrospray ionization mass spectra of representative protein charge ladders of Fe3+Mb, Fe2+-Cyt c, and Cu2+-Az.

12

Figure S2. (a-c) Tandem mass spectra of representative tryptic peptides of acetylated proteins (here, Mb) to confirm the random acetylation of lysine residues and N-terminus. Acetylated residues are highlighted in red within each peptide sequence (Xcorr > 6).

13

Figure S3. (a) UV-vis spectra of protein charge ladders of Fe3+-Mb (met-Mb) and Fe2+-Mb (deoxy-Mb), prepared by reduction with 100 molar equivalents of Na2S2O4 (pH 7.4). (b) UV-vis spectra of protein charge ladders of Fe2+-Cyt c and Fe3+-Cyt c, prepared by oxidation with 5 molar equivalents of K3[Fe(CN)6] (pH = 8.3). (c) UV-vis spectra of protein charge ladders of Cu2+-Az and Cu1+-Az, prepared by reduction with 100 molar equivalents of Na2S2O4 (pH 7.4). UV-vis spectra for all replicate CE assays of lysine-acetyl charge ladders of (d) Fe2+-Mb (n = 9), (e) Fe3+-Cyt c (n = 10), and (f) Cu2+-Az (n = 9) taken after the completion of electrophoresis. Note: running buffers were thoroughly deoxygenated via N2 bubbling immediately before electrophoresis.

14

Figure S4. Replicate CE electropherograms of acetylated (a) Fe3+-Mb (n = 6), (b) Fe2+-Mb (n = 6), (c) Fe2+-Cyt c (n = 7), (d) Fe3+-Cyt c (n = 7), (e) Cu2+-Az (n = 7), and (f) Cu1+-Az (n = 7). Values above each peak or “rung” indicate the corresponding acetylation number. Rung (6) in the electropherogram of Fe2+-Cyt c co-migrated with DMF peak (ZAc(6) ≈ 0), and was not included in calculation of Cyt c net charge. 15

Figure S5. Addition of reducing/oxidizing agents causes the broadening of DMF peak and charge ladders. Comparison between (a) width, and (b) retention time of DMF peaks in the absence of Na2S2O4 (i.e., Fe3+-Mb or Cu2+-Az electropherogram) and in the presence of 10 mM Na2S2O4 (i.e., Fe2+-Mb or Cu1+-Az electropherogram). Comparison between (c) width and (d) retention time of DMF peak in the absence of K3[Fe(CN)6] (i.e., Fe2+-Cyt c electropherogram), and in the presence of 5 mM K3[Fe(CN)6] (i.e., Fe3+-Cyt c electropherogram).

16

Figure S6. Illustration of residues and functional groups that undergo large changes in charge (ΔZ) upon reduction of Fe3+-Mb (PDB: 1YMB) with Na2S2O4 at pH 7.4 (according to calculated changes in pKa). Coordinating residues are included for clarity, even though they are not involved in charge regulation.

17

Figure S7. Illustration of residues and functional groups that undergo large changes in charge (ΔZ) upon oxidation of Fe2+-Cyt c (PDB: 1GIW) with K3[Fe(CN)6] (according to calculated changes in pKa). Coordinating residues are included for clarity, even though they are not involved in charge regulation.

18

Figure S8. Illustration of residues and functional groups that undergo significant changes in charge (ΔZ) upon reduction of Cu2+-Az (PDB: 1E5Z) with Na2S2O4 at pH 7.4 (according to calculated changes in pKa). Coordinating residues are included for clarity, even though they are not involved in charge regulation.

19

Figure S9. Correlation plots of pKa and charge (Z) for all ionizable residues and functional groups of Fe3+-Mb and Fe2+-Mb (a-b), Fe3+-Cyt c and Fe2+-Cyt c (c-d), and Cu2+-Az and Cu1+Az (e-f). Values of pKa are determined via numerical solutions of the Poisson-Boltzmann equation using finite-difference methods,[9] using the following crystal structures: 1YMB (Fe3+Mb), 5D5R (Fe2+-Mb), 1AKK (Fe3+-Cyt c), 1GIW (Fe2+-Cyt c), 5AZU (Cu2+-Az), and 1E5Z (Cu1+-Az). Residues with the largest change in charge (i.e., arbitrarily defined as ΔZ > 0.01) are depicted with open red circles. Values of charge are calculated for each functional group using the Henderson-Hasselbalch equation. 20

21

Figure S10. Correlation plot of (a) total reorganization energy (λ) and (b) redox potential (Eº) vs. the calculated change in free energy associated with single ET (ΔΔGz, defined as the ΔΔG of reduction) for six different proteins. Values of λ represent the total reorganization energy for rutheniated variants of each protein and are extracted from references.[25-28] Values of ΔΔGz were determined via one of two methods of numerical solutions of the Poisson-Boltzmann equation using finite-difference methods,[9, 29] using the following crystal structures: 1YMB (Fe3+-Mb), 5D5R (Fe2+-Mb), 1AKK (Fe3+-Cyt c), 1GIW (Fe2+-Cyt c), 5AZU (Cu2+-Az), 1E5Z (Cu1+-Az), 1AXX (Fe3+-Cyt b5), 1AQA (Fe2+-Cyt b5), 5WQQ (Fe33+Fe2+-HiPIP), 5WQR (Fe23+Fe22+HiPIP), 2QBL (Fe3+-Cyt P450), and 2LQD (Fe2+-Cyt P450). Method 1,[9] described in the text, was used for Mb, Cyt c, and Az. Method 2[29] was used for Cyt b5, HiPIP, and Cyt P450. Method 2 was run on the PDB2PQR server and was less computationally intensive, therefore it was used as a faster alternative for the proteins not examined with CE. Redox potentials were collected from various references.[30-35]

22

Figure S11. Plots of protein dielectric constant versus (a) the calculated change in net charge upon ET (ΔZ); (b) the calculated free energy change associated with single ET (ΔΔGZ); (c) the calculated free energy for the oxidized protein (ΔG); and (d) the calculated free energy for the reduced protein (ΔG). Calculations were performed with the protein dielectric constant values of 1, 2, 3, 4, 5, 10, 20, 30, 40, 60, and 80. The variance of the 0.1 ns snapshots of pKa are indicated with error bars (reported as standard deviation).

23

Supporting Tables: Table S1. Values of pKa and charge for all positively charged amino acid residues in Fe3+-Mb and Fe2+-Mb. Residue

pKa(Fe3+)a

pKa(Fe2+)a

Z Fe

3

Z Fe

2

Z  Z Fe  Z Fe 2

3

H24 7.5476 6.8862 0.584168 0.234480 0.349688 H36 7.0564 7.1234 0.311923 0.345938 -0.034016 H48 7.0657 7.2031 0.316547 0.388580 -0.072032 H64 6.3797 6.7957 0.087119 0.199183 -0.112064 H81 7.6921 7.0340 0.662082 0.300933 0.361149 H82 6.8801 6.4889 0.231969 0.109309 0.122659 H97 6.4760 7.0404 0.106438 0.304047 -0.197609 H113 6.7845 7.0159 0.195094 0.292262 -0.097168 H116 7.3727 7.2211 0.484284 0.398438 0.085846 H119 7.3263 7.0235 0.457688 0.295890 0.161798 K16 11.3116 11.6378 0.999877 0.999942 -0.000065 K42 11.1996 11.4374 0.999841 0.999908 -0.000067 K45 12.0118 12.7770 0.999976 0.999996 -0.000020 K47 11.5652 11.7817 0.999932 0.999958 -0.000027 K50 11.6396 11.4379 0.999942 0.999908 0.000034 K56 11.5307 11.5769 0.999926 0.999933 -0.000007 K62 11.5340 12.1182 0.999927 0.999981 -0.000054 K63 11.0139 11.1896 0.999757 0.999838 -0.000081 K77 11.7851 12.0754 0.999959 0.999979 -0.000020 K78 11.1932 11.6765 0.999839 0.999947 -0.000108 K79 11.2056 11.6064 0.999844 0.999938 -0.000094 K87 10.9448 11.0052 0.999715 0.999752 -0.000037 K96 10.5749 11.1489 0.999332 0.999822 -0.000490 K98 10.8001 11.0185 0.999602 0.999759 -0.000157 K102 11.7800 12.0340 0.999958 0.999977 -0.000018 K118 12.1902 12.4624 0.999984 0.999991 -0.000008 K133 11.9718 12.0135 0.999973 0.999976 -0.000002 K145 12.8166 12.4897 0.999996 0.999992 0.000004 K147 11.2890 11.1135 0.999871 0.999807 0.000064 R31 14.8832 14.6081 1.000000 1.000000 0.000000003 R139 14.7577 15.2279 1.000000 1.000000 0.000000003 NT-G1 8.3735 8.5375 0.903925 0.932079 -0.028155 a Values of pKa for each residue are determined via numerical solutions of the PoissonBoltzmann equation using finite-difference methods,[9] using the following crystal structures: 1YMB (Fe3+-Mb) and 5D5R (Fe2+-Mb).

24

Table S2. Values of pKa and charge for all negatively charged amino acid residues in Fe3+-Mb and Fe2+-Mb. Residue

pKa(Fe3+)a

pKa(Fe2+)a

Z Fe

3

Z Fe

2

Z  Z Fe  Z Fe 2

3

D4 3.2978 3.2226 -0.999921 -0.999934 0.000013 D20 3.2743 3.7134 -0.999925 -0.999794 -0.000131 D44 3.2621 3.6240 -0.999927 -0.999833 -0.000095 D60 3.1865 3.3581 -0.999939 -0.999909 -0.000030 D109 3.9870 4.1424 -0.999614 -0.999448 -0.000166 D122 3.5833 3.3614 -0.999848 -0.999909 0.000061 D126 3.8011 3.8295 -0.999748 -0.999731 -0.000017 D141 3.9700 4.3023 -0.999629 -0.999202 -0.000427 E6 3.4007 3.4099 -0.999900 -0.999898 -0.000002 E18 3.4891 3.6953 -0.999877 -0.999803 -0.000075 E27 3.6422 3.5572 -0.999825 -0.999856 0.000031 E38 3.7217 3.9012 -0.999790 -0.999683 -0.000107 E41 3.9760 4.1735 -0.999623 -0.999407 -0.000217 E52 3.7993 4.0719 -0.999749 -0.999530 -0.000219 E54 3.6465 4.0248 -0.999824 -0.999579 -0.000245 E59 3.8215 3.8341 -0.999736 -0.999728 -0.000008 E83 3.4538 3.6135 -0.999887 -0.999837 -0.000050 E85 3.6595 3.7385 -0.999818 -0.999782 -0.000036 E105 3.0575 3.2905 -0.999955 -0.999922 -0.000032 E136 3.7049 4.0316 -0.999798 -0.999572 -0.000226 E148 3.5857 3.5571 -0.999847 -0.999856 0.000010 CT-G152 2.7640 3.3930 -0.999977 -0.999902 -0.000075 Y103 10.6938 11.0643 -0.000508 -0.000217 -0.000292 Y146 11.1721 11.7692 -0.000169 -0.000043 -0.000126 PropA 3.6958 3.9448 -0.999802 -0.999650 -0.000153 PropD 3.3466 3.4120 -0.999912 -0.999897 -0.000014 – H2O/OH 8.90 15.40 -0.030653 -0.00000001 0.030653 a Values of pKa for each residue are determined via numerical solutions of the PoissonBoltzmann equation using finite-difference methods,[9] using the following crystal structures: 1YMB (Fe3+-Mb) and 5D5R (Fe2+-Mb).

25

Table S3. Values of pKa and charge for all positively charged amino acid residues in Fe3+-Cyt c and Fe2+-Cyt c. Residue

pKa(Fe3+)a

pKa(Fe2+)a

Z Fe

3

Z Fe

2

Z  Z Fe  Z Fe 2

3

H26 5.3929 5.5008 0.009742 0.012455 -0.002713 H33 6.0431 6.2876 0.042113 0.071665 -0.029552 K5 10.4322 10.8109 0.999072 0.999612 -0.000540 K7 9.9161 9.9336 0.996962 0.997082 -0.000120 K8 9.3756 9.6149 0.989533 0.993940 -0.004407 K13 9.0000 9.3146 0.975497 0.987973 -0.012477 K22 11.1578 10.9359 0.999825 0.999709 0.000116 K25 9.6261 10.2331 0.994094 0.998534 -0.004440 K27 9.5191 9.5792 0.992456 0.993424 -0.000969 K39 10.1909 10.0125 0.998384 0.997565 0.000819 K53 10.2107 11.1272 0.998456 0.999813 -0.001357 K55 9.8321 9.8564 0.996316 0.996516 -0.000200 K60 9.8429 10.0354 0.996406 0.997690 -0.001284 K72 9.2720 9.3866 0.986750 0.989792 -0.003042 K73 9.5684 9.4222 0.993260 0.990588 0.002672 K79 9.6424 9.9673 0.994310 0.997299 -0.002989 K86 9.1818 9.2540 0.983741 0.986197 -0.002456 K87 10.5494 9.5591 0.999292 0.993115 0.006177 K88 9.5911 10.2783 0.993601 0.998678 -0.005077 K99 10.6403 10.9245 0.999425 0.999701 -0.000276 K100 10.0859 10.6756 0.997943 0.999470 -0.001527 R38 13.5284 14.5522 0.999999 1.000000 -0.000001 R91 13.7566 13.9182 1.000000 1.000000 0.0000001 NT-G1 8.4462 9.2328 0.917508 0.985517 -0.068008 a Values of pKa for each residue are determined via numerical solutions of the PoissonBoltzmann equation using finite-difference methods,[9] using the following crystal structures: 1AKK (Fe3+-Cyt c) and 1GIW (Fe2+-Cyt c).

26

Table S4. Values of pKa and charge for all negatively charged amino acid residues in Fe3+-Cyt c and Fe2+-Cyt c. Residue

pKa(Fe3+)a

pKa(Fe2+)a

Z Fe

3

Z Fe

2

Z  Z Fe  Z Fe 2

3

D2 1.8120 2.1467 -0.999997 -0.999994 -0.000003 D50 2.3772 2.1588 -0.999991 -0.999994 0.000004 D93 1.6051 2.0370 -0.999998 -0.999996 -0.000003 E4 2.3744 2.4336 -0.999991 -0.999989 -0.000001 E21 2.6769 2.3048 -0.999981 -0.999992 0.000011 E61 1.7593 2.1600 -0.999998 -0.999994 -0.000003 E62 2.8333 3.0981 -0.999973 -0.999950 -0.000023 E66 3.4118 3.4510 -0.999897 -0.999888 -0.000010 E69 2.5043 1.9484 -0.999987 -0.999996 0.000009 E90 1.7463 2.1223 -0.999998 -0.999995 -0.000003 E92 2.5072 2.5201 -0.999987 -0.999987 -0.0000004 CT-E104 1.8588 1.7788 -0.999997 -0.999998 0.0000005 Y48 9.0098 10.8973 -0.023970 -0.000318 -0.023652 Y67 9.9660 10.6346 -0.002709 -0.000582 -0.002127 Y74 9.6878 9.8258 -0.005128 -0.003737 -0.001391 Y97 8.7533 8.9459 -0.042448 -0.027664 -0.014784 Prop6 1.1806 0.8139 -0.999999 -1.000000 0.0000003 Prop7 1.0120 0.6956 -1.000000 -1.000000 0.0000002 a Values of pKa for each residue are determined via numerical solutions of the PoissonBoltzmann equation using finite-difference methods,[9] using the following crystal structures: 1AKK (Fe3+-Cyt c) and 1GIW (Fe2+-Cyt c).

27

Table S5. Values of pKa and charge for all positively charged amino acid residues in Cu2+-Az and Cu1+-Az. Residue

pKa(Cu2+)a

pKa(Cu1+)a

ZCu

2

ZCu

1

Z  ZCu  ZCu 2

1

H35 6.0107 6.7339 0.039207 0.177446 -0.138239 H83 7.7366 8.1348 0.684612 0.844481 -0.159869 K24 13.0453 12.8652 0.999998 0.999997 0.000001 K27 11.1860 11.1629 0.999836 0.999827 0.000009 K41 11.8555 11.6571 0.999965 0.999945 0.000020 K70 11.2900 12.3590 0.999871 0.999989 -0.000118 K74 11.6283 11.8239 0.999941 0.999962 -0.000021 K85 12.2688 12.5090 0.999986 0.999992 -0.000006 K92 10.9641 11.1187 0.999727 0.999809 -0.000082 K101 11.4394 11.5454 0.999909 0.999928 -0.000020 K103 11.5368 11.8765 0.999927 0.999967 -0.000040 K122 10.8403 10.9943 0.999637 0.999746 -0.000108 R79 16.2774 16.4586 1.000000 1.000000 0.0000000005 NT-A1 8.5496 8.7107 0.933832 0.953378 -0.019546 a Values of pKa for each residue are determined via numerical solutions of the PoissonBoltzmann equation using finite-difference methods,[9] using the following crystal structures: 5AZU (Cu2+-Az), and 1E5Z (Cu1+-Az).

28

Table S6. Values of pKa and charge for all negatively charged amino acid residues in Cu2+-Az and Cu1+-Az. Residue

pKa(Cu2+)a

pKa(Cu1+)a

ZCu

2

ZCu

1

Z  ZCu  ZCu 2

1

D6 3.4581 3.6519 -0.999886 -0.999821 -0.000064 D11 3.5932 3.6767 -0.999844 -0.999811 -0.000033 D23 3.1694 3.3711 -0.999941 -0.999906 -0.000035 D55 3.4466 3.6454 -0.999889 -0.999824 -0.000065 D62 2.9398 3.2443 -0.999965 -0.999930 -0.000035 D69 3.6790 3.6277 -0.999810 -0.999831 0.000021 D71 3.6723 4.1968 -0.999813 -0.999374 -0.000439 D76 4.0465 4.2833 -0.999557 -0.999236 -0.000321 D77 2.7718 3.0653 -0.999976 -0.999954 -0.000023 D93 3.1961 3.5980 -0.999937 -0.999842 -0.000095 D98 3.1250 3.2667 -0.999947 -0.999926 -0.000020 E2 3.9989 4.2064 -0.999603 -0.999360 -0.000243 E91 3.5844 3.7341 -0.999847 -0.999784 -0.000063 E104 4.1360 4.1528 -0.999456 -0.999434 -0.000022 E106 4.1781 4.0247 -0.999400 -0.999579 0.000178 Y72 10.1527 10.4814 -0.001764 -0.000828 -0.000936 Y108 10.8946 11.4431 -0.000320 -0.000091 -0.000230 CT-K128 3.3225 3.4758 -0.999916 -0.999881 -0.000035 a Values of pKa for each residue are determined via numerical solutions of the Poisson-Boltzmann equation using finite-difference methods,[9] using the following crystal structures: 5AZU (Cu2+Az), and 1E5Z (Cu1+-Az).

Supporting References: [1] Y. Shi, A. Abdolvahabi, B. F. Shaw, Protein Sci. 2014, 23, 1417. [2] E. Córdova, J. Gao, G. M. Whitesides, Anal. Chem. 1997, 69, 1370. [3] I. Gitlin, M. Mayer, G. M. Whitesides, J. Phys. Chem. B 2003, 107, 1466. [4] B. R. Brooks, C. L. Brooks, A. D. Mackerell, L. Nilsson, R. J. Petrella, B. Roux, Y. Won, G. Archontis, C. Bartels, S. Boresch, A. Caflisch, L. Caves, Q. Cui, A. R. Dinner, M. Feig, S. Fischer, J. Gao, M. Hodoscek, W. Im, K. Kuczera, T. Lazaridis, J. Ma, V. Ovchinnikov, E. Paci, R. W. Pastor, C. B. Post, J. Z. Pu, M. Schaefer, B. Tidor, R. M. Venable, H. L. Woodcock, X. Wu, W. Yang, D. M. York, M. Karplus, J. Comput. Chem. 2009, 30, 1545. [5] A. D. MacKerell, D. Bashford, M. Bellott, R. L. Dunbrack, J. D. Evanseck, M. J. Field, S. Fischer, J. Gao, H. Guo, S. Ha, D. Joseph-McCarthy, L. Kuchnir, K. Kuczera, F. T. K. Lau, C. Mattos, S. Michnick, T. Ngo, D. T. Nguyen, B. Prodhom, W. E. Reiher, B. Roux, M. Schlenkrich, J. C. Smith, R. Stote, J. Straub, M. Watanabe, J. Wiórkiewicz-Kuczera, D. Yin, M. Karplus, J. Phys. Chem. B 1998, 102, 3586. [6] L. Banci, G. Gori-Savellini, P. Turano, Eur. J. Biochem. 1997, 249, 716. 29

[7] V. Rajapandian, V. Hakkim, V. Subramanian, J. Phys. Chem. B 2010, 114, 8474. [8] W. L. Jorgensen, J. Chandrasekhar, J. D. Madura, R. W. Impey, M. L. Klein, J. Chem. Phys. 1983, 79, 926. [9] W. Im, D. Beglov, B. Roux, Comput. Phys. Commun. 1998, 111, 59. [10] M. Nina, D. Beglov, B. Roux, J. Phys. Chem. B 1997, 101, 5239. [11] M. A. S. Hass, F. A. A. Mulder, Annu. Rev. Biophys. 2015, 44, 53. [12] D. Keilin, E. C. Slater, Br. Med. Bull. 1953, 9, 89. [13] M. P. Richards, Antiox. Redox Signal. 2013, 18, 2342. [14] E. Antonini, M. Brunori, Hemoglobin and myoglobin in their reactions with ligands, North-Holland Pub. Co., Amsterdam, 1971. [15] S. Ghosal, Z. Chen, J. Fluid Mech. 2012, 697, 436. [16] B. Gas, M. Stedrý, E. Kenndler, Electrophoresis 1997, 18, 2123. [17] I. De Baere, M. F. Perutz, L. Kiger, M. C. Marden, C. Poyart, Proc. Natl. Acad. Sci. 1994, 91, 1594. [18] D. Bashford, D. A. Case, C. Dalvit, L. Tennant, P. E. Wright, Biochemistry 1993, 32, 8045. [19] M. J. Cocco, Y. H. Kao, A. T. Phillips, J. T. Lecomte, Biochemistry 1992, 31, 6481. [20] R. M. Esquerra, R. A. Jensen, S. Bhaskaran, M. L. Pillsbury, J. L. Mendoza, B. W. Lintner, D. S. Kliger, R. A. Goldbeck, J. Biol. Chem. 2008, 283, 14165. [21] R. Varadarajan, D. G. Lambright, S. G. Boxer, Biochemistry 1989, 28, 3771. [22] T. Simonson, C. L. Brooks, J. Am. Chem. Soc. 1996, 118, 8452. [23] L. Li, C. Li, Z. Zhang, E. Alexov, J. Chem. Theory Comput. 2013, 9, 2126. [24] K. A. Sharp, Biophys. J. 1998, 74, 1241. [25] J. R. Winkler, H. B. Gray, Chem. Rev. 1992, 92, 369. [26] E. Babini, I. Bertini, M. Borsari, F. Capozzi, C. Luchinat, X. Zhang, G. L. C. Moura, I. V. Kurnikov, D. N. Beratan, A. Ponce, A. J. Di Bilio, J. R. Winkler, H. B. Gray, J. Am. Chem. Soc. 2000, 122, 4532. [27] J. R. Scott, M. McLean, S. G. Sligar, B. Durham, F. Millett, J. Am. Chem. Soc. 1994, 116, 7356. [28] M. E. Ener, H. B. Gray, J. R. Winkler, Biochemistry 2017, 56, 3531. [29] T. J. Dolinsky, J. E. Nielsen, J. A. McCammon, N. A. Baker, Nucleic Acids Res. 2004, 32, W665. [30] G. Battistuzzi, M. Borsari, M. Sola, F. Francia, Biochemistry 1997, 36, 16247. [31] J. F. Taylor, V. E. Morgan, J. Biol. Chem. 1942, 144, 15. [32] D. K. Garner, M. D. Vaughan, H. J. Hwang, M. G. Savelieff, S. M. Berry, J. F. Honek, Y. Lu, J. Am. Chem. Soc. 2006, 128, 15608. [33] S. Todorovic, C. Jung, P. Hildebrandt, D. H. Murgida, J. Biol. Inorg. Chem. 2006, 11, 119. [34] M. Rivera, R. Seetharaman, D. Girdhar, M. Wirtz, X. Zhang, X. Wang, S. White, Biochemistry 1998, 37, 1485. [35] L. Liu, T. Nogi, M. Kobayashi, T. Nozawa, K. Miki, Acta Crystallogr. D Biol. Crystallogr. 2002, 58, 1085.

30